futuregraspllc

Diplomacy Turmoil During and Post-COVID19

Thu, 11 Jun 2020 18:12:19 GMT

The coronavirus pandemic has utterly transformed international diplomacy. Statecraft has long been driven by face-to-face interactions that expand through informal mechanisms, such as body language, spontaneous conversations, and an understanding of the views of other actors beyond rote statements of positions. Such meetings are now difficult if not impossible in the era of social distancing and work from home mandates. Meetings postponed or canceled altogether have slowed a range of negotiations while many important international conferences have also been delayed or gone virtual. Cybersecurity risks have amplified as unsecure networks are leveraged for conversations that would have normally occurred face to face. Diplomats must find new ways to work in this pandemic era, as well as what the current turmoil in operations may mean post-COVID19.

The practice of diplomacy has always been intertwined with state-of-the-art methods of information and communications technology (ICT). For example, telegraphy revolutionized the practices of diplomacy by accelerating the speed of communication between embassies and the metropole. The internet has had similar revolutionary impact on the practices of statecraft and diplomacy. Web-enabled ICT multiplied the number of voices and interests involved in international policymaking; complicated international decision-making; reduced exclusive control of States in the process; accelerated and freed the dissemination of information, accurate or not, about any issue or event; and enabled traditional diplomatic services to be delivered faster and more cost-effectively, both to ones’ own citizens and government, and to those of other countries. Following the development of social media, multiple government agencies are now avid posters on Twitter and other platforms in their attempts to shape public discourse at home and abroad.

_________

TURN TO TURMOIL

The novel coronavirus has imposed suddenly a full-court-press for diplomats to take to internet ICT, but accompanying this rapid change are new communications stresses and nuances. The functionalities and efficiencies of diplomatic ICT - including phone, email, social media messaging, and virtual meetings – are now routinely in question. The limitations of video-conferencing and other ICT-driven modes of communication have long been documented . U.S. Foreign Service Officers (FSOs) are specifically trained in face-to-face interactions and talking to groups. Exclusive reliance upon teleconference platforms such as Zoom and WebEx affect multiple core capabilities of the FSO. It is difficult to detect and project desired body language, thus introducing challenges in convincing foreign interlocutors to share private thoughts and intentions over a monitor. Negotiations are both enabled and constrained online, thus pushing the dialogue in sometimes unexpected directions. Reduced or no travel, fewer in-person meetings, and overall, limited personal interactions compromise statecraft. Video engagement changes how information and intelligence is collected, thus impacting institutional memory.

Diplomatic activity and initiatives have slowed considerably as many major meetings and conferences have gone online, with others outright canceled. US Government agencies have suspended many pre-pandemic priorities and pivoted almost entirely to focusing now on coping with the virus. Security in the White House has increased with individuals working near POTUS being given daily coronavirus checks , thus self-limiting the frequency of high-level meetings. Intelligence officers at the CIA are working shifts , such as three days on followed by three days off, to improve social distancing by reducing the given number of people in buildings at any given time. In New York City, itself a pandemic epicenter, all aspects of life have been severely compromised for over a month now, detrimentally affecting United Nations Headquarters operations and activities. For example, COVID-19 has prompted changes in the working methods of the UN General Assembly : “Physical distancing and stay-at-home restrictions mean representatives can no longer meet in person, including to vote on resolutions, which are now circulated under a process known as ‘silence procedure.’ Ambassadors are given a 72-hour window to consult their capitals. If they all agree on a resolution, it is passed. If not, the resolution is not adopted as the ‘silence’ has been broken.”

Webinars and online meetings are the new norm, but they impact the large-scale and often slow-moving collaborative work of drafting and editing of diplomatic agreements and statements. These processes now often involve dozens of people all working remotely using a "track changes" mode of interaction. But direct contact and interaction often undergird these collaborations, as interlocutors know each other and can converse face to face before they move into the web environment. Going straight to the web without initial constructive in-person discussions in the best case merely compromises the speed of production; in the worst case it stymies it entirely.

Cybersecurity risks increase when executing all diplomacy online. A singular reliance on ICT raises new questions about the reliability and vulnerability of networks to disruption or espionage; unpreparedness opens new cyber-attack vectors for hackers, both amateurs and State-sponsored professionals. Governments may not have the requisite cyberinfrastructure and trained employees to seamlessly continue diplomacy operations from home offices. Sharing of sensitive documents either cannot be done at all, or only in a limited manner. Diplomatic online sessions can be recorded and played back with deep analyses of verbatim words and body language – normally hurried notes taking and brief summarizations are what constitute the record for face-to-face conversations. Since different organizations and different countries can adopt disparate and non-interoperable technological approaches, ICT compatibility confusion throws additional barriers at smooth statecraft. Some diplomats may avoid negotiations altogether because of security concerns, thus compromising statecraft progress in general.

_____________

ACCESSING THE ADVANTAGE

State engagements must continue apace regardless of the pandemic. Negotiations of peace, non-virus global health issues, human rights, climate change, nuclear nonproliferation, and trade all must occur, in whatever limited way. Embassies need to find ways to provide emergency consular services to visiting or expatriate citizens in need without putting diplomats at risk of infection. Diplomats face challenges in assessing emerging political, economic, and social trends when prevented from face-to-face meetings or travel around a country. National elections scheduled during this period of social distancing may take place without the confidence-building presence of foreign election observers, increasing potential for corrupt or anti-democratic practices to alter election outcomes.

Diplomatic professionals will have to consider how the new online reality impacts their ability to advance their national interests, to understand rapidly shifting geopolitical events, and to reach publics overseas. Key is to understand how diplomacy morphs into a new normal in the coming months and years, and to identify means that diplomats and senior policymakers can prepare. Tactical challenges are to identify best ways of doing business and to implement them in real time during the pandemic. Strategic challenges are to consider how diplomacy following the pandemic might inalterably change, what new opportunities are present, and how to position diplomats accordingly.

Advantages and opportunities exist in this new online reality, nevertheless. Frictionless meeting logistics with accompanying financial savings for internet-only engagements; increased opportunities to hear directly from key players; younger diplomats shining with their greater digital skills – all these and more are coming to light in the pandemic era. As stated by the United Nations’ Under-Secretary General for Political and Peacebuilding Affairs, Rosemary DiCarlo , “Although we recognize that the limitations of the processes in which face-to-face meetings are restricted, the increased use of technology has the potential to create new opportunities [and] enhance the inclusivity of peace processes - for example, including the participation of women and young people.”

Concomitant with this new reality is the increased potential to leverage big data, predictive analytics, and other digital technologies. Investments should be amplified in artificial intelligence (AI) to analyze all data that can be extracted on any given country, region, or topic. Predictive analytics using machine learning (and its subset, deep learning) can then be applied to offer novel insights for meeting preparedness and briefings to senior policymakers; such approaches offer new opportunities in optimized human-machine hybrid decision-making. AI can also be used to automatically assess treaty compliance, as well as to monitor and predict events before they occur . In-situ machine language translations and real-time content assessment for both verbal and written communiques are feasible now. Increased security of voting processes via blockchain holds great potential. Virtual reality avatars can be applied to replicate the feeling of sitting across a table from a counterpart. Ultimately, there is a wide range of technological applications and solutions that await adoption and deployment by forward-leaning diplomatic corps.

However long the pandemic lasts, diplomacy has been forever changed. How diplomats respond to the recent step-change in personal interactions and resultant altered statecraft will dictate whether national objectives are met, whether treaties are negotiated, and whether normal operations return - or diplomacy turmoil continues.

The opinions and characterizations in this piece are those of the authors and do not necessarily represent those of the U.S. Government.

Thomas A. Campbell, Ph.D. (tom.campbell@futuregrasp.com) is Founder & CEO of FutureGrasp, an advisory group that works with organizations globally to identify policy and business opportunities from emerging and disruptive technologies, especially artificial intelligence. Previously, he was the first National Intelligence Officer for Technology with the National Intelligence Council, Office of the Director of National Intelligence (ODNI).

Andrew Hyde (andrew.hyde@futuregrasp.com) is a Nonresident Fellow at the Stimson Center, a Washington, D.C. based think tank, with the Transforming Conflict and Governance Program, and a Senior Advisor with FutureGrasp. Previously, he was a Foreign Service Officer at the U.S. Department of State and a Congressional staffer.

Geoffrey Odlum (geoffrey.odlum@futuregrasp.com) served as a Foreign Service Officer at the U.S. Department of State from 1989 to 2017. He is currently the President of Odlum Global Strategies, a national security and technology policy consulting firm, and is a Senior Advisor with FutureGrasp.

Ariel I. Ahram, Ph.D. (ahram@vt.edu) is Associate Professor and Chair of Government and International Affairs (GIA) at the School of Public and International Affairs, Virginia Tech.

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

AI National Plans: Their Necessity and Core Components

Thu, 04 Apr 2019 20:16:29 GMT

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Artificial intelligence (AI) is one technology sector where strategic planning to guide diverse resources is especially valuable for governments. The consulting firm PwC estimates that AI will add $15.7 trillion to the world economy by 2030 [1] ; such eye-popping numbers don’t go unnoticed by senior policymakers of major governments. Consequently, several nation states are taking the initiative to get organized so that they may capitalize upon the algorithms that are driving the Fourth Industrial Revolution. [2] In addition to sponsoring large (in many nation states, multi-billion US dollar) R&D investments, a means toward this facilitation is to draft and execute AI national plans. Here we address the global need for such strategic documents, as well as the core aspects of them required for maximum effectiveness.

At this writing, only a few dozen nation states have published or have begun to draft AI national plans. A report soon to be published from the United Nations Interregional Crime & Justice Research Institute (UNICRI) and FutureGrasp, LLC [3] details that only 14 nation states have active AI national plans. Moreover, only 19 have expressed interest in developing one (as demonstrated by a working group formed and/or an outline or a white paper published toward an AI national plan). These are surprisingly low numbers, particularly considering the multi-decadal history of AI and the total of 193 sovereign states recognized by the United Nations.

Without a guiding document that encompasses the unique culture and capabilities of each nation state, there is the risk that AI will soon become beyond the direct control of some nation states. Given the deep infusion of AI into essentially every economic sector (including healthcare, finance, transportation and many others), inactive nation states will become reliant upon other countries, such as China, for their current and future computation. As every year goes by, the ability to catch up to those nation states that have active AI national plans will become increasingly difficult. Even the United States - which presently lacks an AI national plan, yet still leads AI overall because of its industry and academic domination - is not immune to being surpassed by other nation states that are more proactive in mustering their national capabilities. [4] [5]

It is the position of FutureGrasp, LLC that every nation state should have an AI national plan. To remain economically competitive, as well as to facilitate collaboration across borders, it is imperative that governments recognize the need to channel whatever resources they possess to organize strategic directions in the rapidly growing technology sector of AI.

Toward that end, here we detail a few core aspects of AI national plans that nation states should consider as they grapple with how to manage all the facets needed within the cross-cutting platform capabilities of AI. We also offer an example in each AI sector of a nation state that has considered it. From the earlier-mentioned UNICRI-FutureGrasp report [ 3] , the AI sectors of core interest to any successful AI national plan are: Research, Infrastructure, Talent/Education, Industrial Strategy, Ethics/Legal Implications, AI in Government, and Data.

Research. Supporting a robust AI ecosystem is perhaps the most salient of all AI sectors. Like many emerging technologies, AI is in a dual-nature state of both current applications (for example, deep learning for facial recognition) and forward-leaning research (for example, advanced algorithms for autonomous vehicles). The only way to maintain an edge in AI is to fund basic research. While industry does sponsor significant research, such funding is driven ultimately by quarterly profits. Truly advanced research must be funded by governments. For example, DARPA [6] in the United States announced in September 2018 that they would fund US $2 billion to seek “contextual reasoning in AI systems to create more trusting, collaborative partnerships between humans and machines.” [7] While many nation states cannot afford US $2 billion in investments, even much smaller amounts can help advance the state of AI within a respective country. One strategic approach is to target a single niche area where a country has specific expertise and seek to become the world’s best at it; for example, AI analysis of video from autonomous vehicles. Funding new research centers, hubs and programs in basic and applied AI benefits nation states immediately with new employment, and long-term with new AI capabilities that can yield gains in government and the private sector.

Infrastructure. AI is part of the digital ecosystem that surrounds us all. Beyond mere algorithms, AI requires the development of computational resources (semiconductors, enterprise systems, cloud computing), sensors (Internet of Things), and wireless services (3G, 4G and LTE now, 5G soon). AI infrastructure complements other needed national plans in those related technical spaces. Without this digital infrastructure, any AI national plan is doomed to be incomplete – having great algorithms is insufficient if one cannot provide data to them, compute them, and translate their results into meaningful output. A stellar example of a city-in-progress by Saudi Arabia that considers digital infrastructure to its fullest is that of NEOM, a US $500 billion initiative. [8] NEOM is designed to provide all the latest technologies (5G, Internet of Things, advanced compute), while simultaneously leveraging the full capabilities of applied AI.

Talent/Education. People are what makes technology tick. Necessary for a successful AI national plan are core aspects of the talent/education AI sector: a robust education system (particularly in STEM – science, technology, engineering and mathematics); welcoming immigration policies for people with top AI talents; funding for university scholarships and faculty chair positions; promoting diversity among AI researchers; and lifelong learning and retraining programs. An example of commitment from a nation state in their AI national plan toward talent/education is the United Kingdom. Within their plan, they devote several paragraphs to “Work with schools, universities and industry to ensure a highly skilled workforce; Enable access to high-skilled global talent; and Take steps to promote diversity in the development of AI.” [9] Commensurate funding of hundreds of million British pounds (£) is assigned to support this critical AI sector.

Industrial Strategy. The transition of government-funded R&D into the private sector is a critical component of any AI national plan. This handoff can be catalyzed by funding to AI startups and promotion of innovation ecosystems (for example, industrial parks). An example of successful industrial strategy execution is Canada. Within its Pan-Canadian Artificial Intelligence Strategy, [10] Canada highlights industry engagement as a core component of its AI national plan. Moreover, Canada backed up their words with funding: “Canada did a lot in 2017 to build on our AI research advantage with more than CAN $300-million [US $229 million] in new funding for research and $260-million [US $199 million] raised by Canadian AI startups.” [11]

Ethics/Legal Implications. AI has potential to create disruption in the legal community, especially in the context of ethics implications. For example, if an autonomous vehicle causes an accident, who is legally responsible: the programmer? the car manufacturer? the passenger? Ethics also enter prominently into such discussions. If an autonomous vehicle has to make a split-second decision whether to swerve and crash into a telephone pole (thus potentially killing the driver) or to steer toward and potentially kill a pedestrian, which is the correct ethical choice? The permutations of such debates rapidly increase with more people involved and more complex systems. Rather than leaving ethics after incidents to the philosophers and legal issues to the lawyers, it behooves governments to infuse such issues into their AI national plans. Regulations for the application of AI can then be established that benefit society as a whole. The South Korean AI national plan [12] is an exemplar of tackling this issue. Within that plan is mandated the task to “Establish a charter of ethics for Intelligent IT to minimize any potential abuse or misuse of advanced technology by presenting a clear ethical guide for developers and users alike (2018).”[ 12]

AI in Government. Governments are often the most complex organizations and biggest employers within a nation state. It makes sense then to leverage applied AI within government to increase efficiencies, enhance communications to the public, and manage government agencies in general. Italy may be unique among nation states with the initial focus in its AI national plan upon how its government might leverage AI in its public administration. The Italian Government, led by a task force [13] , released a white paper on AI in March 2018, Artificial Intelligence: At the Service of Citizens. “The objective is to facilitate the adoption of these technologies in the Italian Public Administration, to improve services to citizens and businesses, thus giving a decisive impulse to innovation, the proper functioning of the economy and, more generally, to progress in daily life.” [14]

Data. Machine learning and deep learning algorithms require data, and lots of it. Fortunately, governments have incredible volumes of data: tax records, health information, economic trends, census data, and many more 1s and 0s fill government computers. The challenge is to harness that information in a meaningful manner so AI algorithms can apply it, while still adhering to privacy constraints and other data policies. With its recent AI national plan, France presents a good example of how to develop data policies for applied AI. Of the six parts of its plan, Part I is titled “An Economic Policy Based on Data.” Within that section is detailed how France will leverage its strengths in data management while keeping in mind European policies, such as the GDPR. [15] To avoid “spreading efforts too thinly,” the French Government proposes to focus upon only four sectors initially: healthcare, environment, transport mobility and defense-security.

As the field of AI continues to accelerate, nation states that become engaged now will have an advantage long-term compared to inactive nation states. The AI sectors detailed above offer nation states key points to consider as they strategize and enter initially or engage more deeply in the exciting world of AI.

Notes (all weblinks accessed March 2019)

[1] “Sizing the prize - PwC’s Global Artificial Intelligence Study: Exploiting the AI Revolution,” 2017, https://www.pwc.com/gx/en/issues/data-and-analytics/publications/artificial-intelligence-study.html

[2] Klaus Schwab, “The Fourth Industrial Revolution: what it means, how to respond,” January 14, 2016, https://www.weforum.org/agenda/2016/01/the-fourth-industrial-revolution-what-it-means-and-how-to-respond/

[3] United Nations Interregional Crime and Justice Research Institute (UNICRI) & FutureGrasp, LLC; “Artificial Intelligence: State Initiatives & Crime and Justice Implications,” in final edits and reviews.

[4] Kai-Fu Lee, “AI Superpowers: China, Silicon Valley, and the New World Order,” 2018, Houghton Mifflin Harcourt.

[5] Joshua New, December 4, 2018, “Why the United States Needs a National Artificial Intelligence Strategy and What It Should Look Like,” Center for Data Innovation, http://www2.datainnovation.org/2018-national-ai-strategy.pdf

[6] Defense Advanced Research Projects Agency

[7] “DARPA Announces $2 Billion Campaign to Develop Next Wave of AI Technologies,” September 7, 2018, https://www.darpa.mil/news-events/2018-09-07

[8] T. Zareva, October 30, 2017, “Saudi Arabia Is Building a $500-Billion New Territory Based on Tech and Liberal Values,” https://bigthink.com/design-for-good/saudi-arabia-is-building-a-utopian-city-to-herald-the-future-of-human-civillization

[9] “Industrial Strategy; Artificial Intelligence Sector Deal,” https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/702810/180425_BEIS_AI_Sector_Deal__4_.pdf

[10] “CIFAR Pan-Canadian Artificial Intelligence Strategy,” https://www.cifar.ca/ai/pan-canadian-artificial-intelligence-strategy

[11] S. Irvine, “From startups to champions: How to focus Canada’s AI dreams,” February 2, 2018, https://www.theglobeandmail.com/report-on-business/rob-commentary/from-startups-to-champions-how-to-focus-canadas-ai-dreams/article37836853/

[12] “Mid- to Long-Term Master Plan in Preparation for the Intelligent Information Society Managing the Fourth Industrial Revolution,” Government of the Republic of Korea—Interdepartmental Exercise, https://english.msit.go.kr/cms/english/pl/policies2/__icsFiles/afieldfile/2017/07/20/Master%20Plan%20for%20the%20intelligent%20information%20society.pdf

[13] “IA Task Force,” https://ia.italia.it/task-force/

[14] “Artificial Intelligence—At the Service of Citizens,” March 2018, Task Force on Artificial Intelligence of the Agency for Digital Italy, https://ia.italia.it/assets/whitepaper.pdf

[15] General Data Protection Regulation

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Digital Twins and Digital Doubles: Populating our Virtual Worlds

Fri, 07 Sep 2018 14:36:18 GMT

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Digital Twins

Years ago when I worked in the semiconductor crystal growth industry, we had a frequent problem with crystal defects that occurred, seemingly at random, during a multi-day growth process. These defects caused total failure of the greater than $80,000 single crystal. One of my proposed solutions was to create a dynamic computer model of the whole system. The simulation would run simultaneously with the real system’s control settings and model in real-time the various potentials for defects well in advance of their occurring. The closed-loop SCADA [1] system would then make appropriate corrections to avoid defects and yield a perfect crystal. While theoretically possible, my proposal was unfortunately ahead of its time in terms of the compute power and sensors required vs. those available in the early 2000’s.

Fast forward to today, and this concept is called a ‘digital twin,’ defined as “a virtual model that is essentially the intelligence counterpart to an actual, physical object.” [2] Now digital twins are becoming relatively common - “Gartner predicts that by 2021, half of large industrial companies will use digital twins, resulting in those organizations gaining a 10% improvement in effectiveness.” [3]

Digital twins rely on three main components: data (from sensors in the Internet of Things, IoT); a ‘digital thread’ that relays information to the digital twin [4] ; and computation (from artificial intelligence, AI). Fortunately, all three are now up to snuff enough to enable useful applications of digital twins. For example, in the oil and gas industry, predictive maintenance via digital twins offers tremendous benefits:

“The digital twin is a system of systems based on a virtual digital copy of all the infrastructure assets as represented by Deep Learning Neural Networks (DNNs). DNNs are a Machine Learning technique that brings us a new paradigm of ‘software that writes software’ and acts as a compiler for your data to deliver the desired outcome…There is no doubt that smarter, faster, AI-powered applications and digital twins can have a fundamental effect on predictive maintenance in the oil & gas industry, but it will also pave the way for AI and GPU infrastructure enabled edge-to-cloud capabilities to enable full-stream outcomes for production optimization, automated drilling and even bridging the silos across all the operational streams from oil exploration and extraction to processing and distribution.” [5]

While industrial systems may be relatively straightforward to simulate, could we advance this concept dramatically to simulate a living human, thereby creating a ‘digital double’? What benefits and applications might be created if we have a virtual simulacrum of ourselves with all our knowledge, health status, personality quirks, and worldviews? With faster compute, increasing volumes of data, the creation of multi-purpose platform apps, and artificial intelligence (AI), we are getting close to realizing digital doubles. The implications could be profound.

Platform Apps

In the fiction book, “The Circle,” by Dave Eggers [6] , a single company becomes the mashup equivalent of Google, Facebook, Amazon and sundry other companies in its effort to “close the circle” and obtain total digital control and transparency among all global citizens. The book resonated with many people in the technology community, as it is not a far cry from what we see already in China and Silicon Valley.

A close real equivalent to the fictional company Circle is the Chinese firm WeChat (developed by Tencent), which is simultaneously a messaging, social media and mobile payment digital platform. As of March 2018, WeChat has over 1 billion users. Moreover, it is close to being a dominant platform for mobile users – “Chinese users spend approximately one third of all their time on mobile in WeChat.” [7] WeChat is in the vanguard of multi-app consolidation and control.

As with WeChat, the past decade has seen the emergence of platform apps with many capabilities all packaged under a single user-login. Facebook, Amazon, Alibaba and Google are paradigms of what software companies can become through strong programming and rapid absorption of their competition. Apple (iPhone), Samsung (Galaxy), and Google (Android) show what hardware firms can accomplish by stuffing ever more AI processing power into a smartphone. By becoming go-to platform apps, these companies have positioned themselves as five of the top ten – Apple, Amazon, Alphabet, Microsoft and Facebook - in market capitalization listings. [8]

Thus, the trend is clear toward continuing consolidation of capabilities under a handful of umbrella app platforms. The full manifestation of this app-merging would be for each of us to get a digital double – “virtual models of ourselves that will interact with each other in countless simulations to help us make faster, more informed choices in our daily lives.” [9] Such virtual humans would offer significant benefits.

Digital Doubles

In a recent article in Scientific American , Pedro Domingos expounds upon what he sees as a trend in the AI community toward the digital double. Prof. Domingos researches what he dubs the “Master Algorithm,” [10] a fusion of deep learning, symbolic learning, evolutionary algorithms, analogical methods and Bayesian learning. Promises for the Master Algorithm’s benefits are significant:

“Once we attain the Master Algorithm and feed it the vast quantities of data each of us produce, artificial intelligence systems will potentially be able to learn very accurate and detailed models of individual people: our tastes and habits, strengths and weaknesses, memories and aspirations, beliefs and personalities, the people and things we care about, and how we will respond in any given situation. That models of us could essentially predict the choices we will make is both exciting and disquieting.” [ 9]

An important step for digital doubles is to create seamless voice communications. Several forward-thinking digital companies have marketed voice-activated systems. Apple voices Siri; Microsoft offers Cortana; and Amazon gives us Alexa. While still far from perfect, with more data and continuous improvements in compute, the time will come soon where it will be normal to have a conversation with a computer.

But voice is only one component of a digital double—even more critical is the modeling of an actual person. A few companies are leaders in this space. IPsoft offers the digitized human of Amelia, which possesses cognitive learning abilities, autonomic task management, and emotional intelligence. Amelia services a range of industries like banking, insurance and healthcare to remove much of the drudgery of customer assistance. [11] The company Oben “is building a decentralized AI platform for Personal AI (PAI), intelligent 3D avatars that look, sound, and behave like the individual user.” One can then use that avatar to perform tasks that one may not have time for or would rather not execute. [12]

What might be some applications for a digital double? Let’s review some possibilities.

· Communications – Digital doubles would have the capacity to chat with each other. Suppose you’re planning a party and want to invite all your best friends. You tell your digital double to contact all their digital doubles and arrange the date, time, who brings what, etc. All the knowledge of your friends' calendars, likes/dislikes, and anything else that might affect the outcome of the event could be considered in the virtual world and then brought into reality at the big event.

· Travel – Your virtual persona could be the ideal travel agent. It would know what you prefer to do, where you want to visit, your comforts and discomforts, and really anything else that might affect your personal travel. Moreover, it could navigate all those multiple apps for booking your airline flight, hotel room, rental car, tickets, etc. It could even help you take and choose which photos to post later in social media.

· Knowledge Acquisition – Instead of typing into Google to find that random factoid you want to know, if you have a digital double you could just ask them. With full internet access, your digital double would be able to find any piece of information you’d want within milliseconds. Knowing your interests in great detail, your digital double could also suggest additional information beyond the original search.

· Healthcare – A Holy Grail for a digital double would be to have it mimic essentially the entire human body, in all its nuances. This might be the last stage of the development of virtual self-avatars, as the complexity of the body (37.2 trillion cells in the human body [13] , 100 billion neurons in the human brain [14] ) would challenge even the most advanced supercomputers of today. Nevertheless, we could make basic simulations of the body in digital doubles and thus aid doctors in diagnosing diseases, optimizing nutrition, and assisting in exercise.

Conclusions

Industry already leverages digital twins to streamline equipment maintenance, identify potential part failure modes, and automate more of its processes. The capabilities of digital twins will only improve with better sensors, faster cloud and edge computing, and ever-improving AI.

Will society take the next step and simulate actual humans in the form of digital doubles? I believe it is inevitable. Many of us already have a substantial digital presence, whether in Facebook, LinkedIn, Twitter, WeChat or other social media sites. We spend increasing amounts of time online. A next step for our digital lives would be to create digital doubles that know the most intimate details of us. Benefits for those with a constantly learning virtual self-avatar could be substantial.

As we move into the future with more advanced compute and exponential data availability, there will be an inexorable push to simulate more of our world. Populating our digital worlds with digital twins and digital doubles will ultimately serve us greatly in our increasingly complex real worlds.

Notes (all links accessed September 2018)

[1] Supervisory control and data acquisition

[2] “Using AI, IoT and Big Data to Deliver Digital Twins,” September 2, 2017, https://insidebigdata.com/2017/09/02/using-ai-iot-big-data-deliver-digital-twins/

[3] C. Pettey, “Prepare for the Impact of Digital Twins,” September 18, 2017, https://www.gartner.com/smarterwithgartner/prepare-for-the-impact-of-digital-twins/

[4] “Industry 4.0 and the digital twin,” Deloitte Consulting, 2017, https://www2.deloitte.com/content/dam/Deloitte/cn/Documents/cip/deloitte-cn-cip-industry-4-0-digital-twin-technology-en-171215.pdf

[5] P. Modi, “How AI Is Providing Digital Twins For Predictive Maintenance In Oil And Gas,” June 21, 2018, https://www.forbes.com/sites/nvidia/2018/06/21/how-ai-is-providing-digital-twins-for-predictive-maintenance-in-oil-and-gas/

[6] D. Eggers, “The Circle,” 2013, Alfred A. Knopf & McSweeney’s Books, San Francisco, https://www.amazon.com/dp/B00EGMQIJ0/ref=dp-kindle-redirect?_encoding=UTF8&btkr=1

[7] M. Brennan, “One Billion Users And Counting -- What's Behind WeChat's Success?,” March 8, 2018, Forbes, https://www.forbes.com/sites/outofasia/2018/03/08/one-billion-users-and-counting-whats-behind-wechats-success/#57ecfb57771f

[8] “The Largest Companies By Market Cap (100),” http://www.symbolsurfing.com/largest-companies-by-market-capitalization

[9] P. Domingos, “Our Digital Doubles,” Scientific American, pp. 88-93, September 2018.

[10] P. Domingos, “The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World,” 2015, Basic Books, New York, NY, https://www.amazon.com/dp/B012271YB2/ref=dp-kindle-redirect?_encoding=UTF8&btkr=1

[11] IPsoft, https://www.ipsoft.com/amelia/

[12] Oben, https://oben.me/

[13] E. Bianconi, et al., “An estimation of the number of cells in the human body,” July 5, 2013, Annals of Human Biology, https://www.tandfonline.com/doi/abs/10.3109/03014460.2013.807878

[14] “The Human Memory,” http://www.human-memory.net/brain_neurons.html

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Terminator Conundrum: Perhaps a Middle Ground?

Fri, 10 Aug 2018 14:40:56 GMT

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

HISTORY REPEATING?

History is replete with examples of governments that agreed to temporary moratoria or outright bans on technologies, but then proceeded to ignore those restrictions when circumstances changed. In 1899, attendees at the Hague Convention enacted a prohibition on discharging projectiles or explosives from balloons during war. [1] Nation states then proceeded to ignore their signatures when the use of aerial weapons was proved invaluable in World War I. [2]

In the early days of the 20th century, nations banned unrestricted submarine warfare, in which submarines sink civilian vessels such as freighters and tankers without warning. [3] Nevertheless, the Germans ignored these restrictions in World War I, as exemplified by the dramatic sinking of the RMS Lusitania by a German U-boat in 1915. [4] Such actions were condemned worldwide, especially by the United Kingdom and the United States…then Pearl Harbor occurred, and all gloves were off. The order to “EXECUTE AGAINST JAPAN UNRESTRICTED AIR AND SUBMARINE WARFARE” was issued to all US Navy ship and submarine commanders on December 7, 1941, just four and a half hours after the attack at Pearl Harbor (as noted in the excellent book, "Army of None: Autonomous Weapons and the Future of War," by Paul Scharre, [5] ) A massive US submarine campaign in the Asia-Pacific region was initiated to attack Japanese merchant vessels in defense of US troops deploying there. [6]

It is often said that history repeats itself—we seem to be seeing these same trends today with autonomous weapons. During the Obama Administration, Deputy Secretary of Defense, Robert O. Work, heavily promoted what he called the “Third Offset,” which sought to outmatch adversaries completely via advanced systems, including autonomous weapons and artificial intelligence (AI). Despite this desire for battlefield dominance, Mr. Work nevertheless emphasized that there would always be a “human in the loop,” meaning the systems would never become completely autonomous—especially to be able to make kill decisions on their own. [7] However, R&D on such intelligent systems has created significant controversy.

PUSHBACK

Talk of and investment into autonomous weapons has initiated fierce debate within defense and technology circles. What happens if the United States insists on this policy of a human in the loop, while its adversaries remove that step with fully autonomous weapons, and they consequently can act faster and with a higher kill rate on the battlefield? Moreover, are such systems error-free? Bias-free? Reliable? Given these and other technical and ethical concerns, in 2016 the Vice Chairman of the Joint Chiefs of Staff, Gen. Paul J. Selva, labeled the question of whether we should remove human oversight from any automated military operation as the Terminator Conundrum , after the Terminator movies in which robots take over the world. [8]

International organizations have also weighed-in. Since 2013, the United Nations has had an active commission focused on addressing questions regarding lethal autonomous weapons systems (LAWS). [9] Several countries have even gone so far as to support a ban on lethal autonomous weapons, although curiously the list does not include any of major military powers. “Only twenty-two nations have said they support a ban on lethal autonomous weapons: Pakistan, Ecuador, Egypt, the Holy See, Cuba, Ghana, Bolivia, Palestine, Zimbabwe, Algeria, Costa Rica, Mexico, Chile, Nicaragua, Panama, Peru, Argentina, Venezuela, Guatemala, Brazil, Iraq, and Uganda (as of November 2017). None of these states are major military powers and some, such as Costa Rica or the Holy See, lack a military entirely.” [ 5 ]

Fast forward two years, and we see the controversy heating up even more. In June 2018, more than 3,000 Google employees protested Google’s activities with the US Department of Defense (DoD) in a letter to Google’s C-suite (about a dozen employees are said to have even resigned). They demanded that Google cease engagement in the DoD’s Project Maven, which was focused initially upon using AI to improve image analysis. Google’s leadership subsequently announced it would not renew its contract with the DoD. Moreover, Google drafted guidelines for AI that exclude the development of autonomous weapons and most applications of AI that might harm people. [10] In late July 2018, 2,400 researchers in 36 countries joined 160 organizations in signing a letter, drafted by the Future of Humanity Institute at MIT, to “call upon governments and government leaders to create a future with strong international norms, regulations and laws against lethal autonomous weapons.” [11] , [12] , [13] This letter builds upon a similar letter signed in 2015. [14]

While admirable for standing up for their principles or ethics, it is doubtful that these individual protests, corporate withdrawals, or international bodies will sway governments from continuing to develop and deploy autonomous weapons. The advantages of autonomous weapons are just too enticing for some countries, especially those with different ethics and governments than Western countries:

“Robots have many battlefield advantages over traditional human-inhabited vehicles. Unshackled from the physiological limits of humans, uninhabited (“unmanned”) vehicles can be made smaller, lighter, faster, and more maneuverable. They can stay out on the battlefield far beyond the limits of human endurance, for weeks, months, or even years at a time without rest. They can take more risk, opening up tactical opportunities for dangerous or even suicidal missions without risking human lives.” [ 5 ]

Three nation states in particular currently lead efforts toward developing these systems: the United States, China and Russia.

NATION STATE INTEREST

United States. The aforementioned Project Maven is the first wave of a much larger planned AI focus within the DoD. In June 2018, the DoD initiated the Joint Artificial Intelligence Center (JAIC) , with the “overarching goal of accelerating the delivery of AI-enabled capabilities, scaling the Department-wide impact of AI, and synchronizing DoD AI activities to expand Joint Force advantages.” [15] Project Maven and other AI projects will be folded into the JAIC. The DoD requested $75 million to standup JAIC, with an anticipated need of $1.7 billion over six years. [16] Thus, regardless of Google’s and other stances against autonomous weapons, the DoD is proceeding forward with its own research and development. Potential strategic advantages for C4ISR [17] are simply too large to ignore for the world’s largest defense department. Furthermore, given the pots of money involved (“Global spending on military robotics is estimated to reach $7.5 billion per year in 2018, with scores of countries expanding their arsenals of air, ground, and maritime robots.” [ 5 ]), other companies are more than happy to fill the void that Google leaves behind to support the DoD. Meanwhile, US adversaries are doing even more.

China. Previously not engaged seriously in the AI community, China had a wake-up call in 2016 when the AI company DeepMind beat the world-champion Go player, South Korean Lee Sedol, with its algorithm AlphaGo. Long considered a difficult game to master (with essentially infinite move variations compared to the more pedestrian game of chess), Go presented programmers with a seemingly impossible task for challenging world-class human players. Then DeepMind beat Mr. Sedol in 4 games to 1, proving that AI (at least deep learning) has come of age. [18] Some see simulating the game Go as an inchoate step toward full battlefield simulations. [19]

Chinese investment in AI quickly ramped-up. China announced a plan in 2017 to become the world leader in AI by 2030. [20] Its nation state investments go beyond merely algorithms: “Its investment in A.I., chips and electrics cars combined has been estimated at $300 billion.” [21] Taking a holistic view of technologies with the potential to complement and accelerate AI is critical for success in autonomous weapons – areas such as semiconductors, data storage, sensors, cloud computing, robotics, and supercomputers all affect the pace of AI development. [22]

China plans to leverage this funding and AI capabilities to create a suite of autonomous weapons. For example, Chinese industry is testing swarms of UAVs [23] (up to 1,000 at once) in airshows to demonstrate their AI orchestration for advanced maneuvers. There is strong interest by the Chinese military in applying such drones in combat situations. “The attraction of mastering drone swarms resonates in both the defensive and offensive realms of warfare. They can be used as part of air defense measures in the event of an attack and a major war, being operated by another aircraft and released in mid-air by fighter jets. In an offensive scenario, smart swarms can drastically undermine and neutralize enemy defense systems, freezing defensive elements in a given battlefield and allowing for near-free movement and roaming of offensive forces.” [24] Moving from UAVs to UUVs [25] , the PLA [26] recently gave an interview in which they detailed a classified project to arm submarines with AI for “handling such tasks as surveillance, minelaying, and even attacks on enemy vessels, relying on AI to adjust to changing conditions." [27] It remains to be seen how effective these tools are, but we can be assured that China’s interest in AI and autonomous weapons is profound and its advances will continue.

Curiously, at the last UN LAWS meeting (April 2018), China endorsed a call to prohibit the use of autonomous weapons, yet they did not address their development or production. [28] This strikes some as a bit specious, given China’s already large investments into AI. By not endorsing a ban on their development or production, China might actually gain an advantage. They would appear to be playing along with the international community, yet never cease their R&D on autonomous weapons, thus retaining the option to deploy them if their senior leadership felt the need.

Russia. Russia is also aggressively pursuing AI. In a speech to school children in 2017, Vladimir Putin stated, “Artificial intelligence is the future, not only for Russia, but for all humankind. It comes with colossal opportunities, but also threats that are difficult to predict. Whoever becomes the leader in this sphere will become the ruler of the world.” [29] He further commented, “when one party's drones are destroyed by drones of another, it will have no other choice but to surrender." [30] Critically, Russia announced in 2017 that they would not be influenced by the UN bodies seeking for a ban on autonomous weapons. [31]

In July 2018, Russia released an ambitious 10-point plan for organizing its AI efforts, including military engagements. [32] Russia is particularly active in robotics:

“While the United States has been very reluctant to arm ground robots, with only one short-lived effort during the Iraq war and no developmental programs for armed ground robots, Russia has shown no such hesitation. Russia is developing a fleet of ground combat robots for a variety of missions, from protecting critical installations to urban combat. Many of Russia’s ground robots are armed, ranging from small robots to augment infantry troops to robotic tanks. How much autonomy Russia is willing to place into its ground robots will have a profound impact on the future of land warfare.” [ 5 ]

Thus, Russia is not dissuaded by calls for bans on autonomous weapons; on the contrary, it is accelerating their efforts to field these systems.

LIMITATIONS

Despite such strong nation state interest in autonomous weapons, there are still many fundamental limitations AI must overcome before science fiction dreams of weapons developers are achieved. Issues such as algorithmic bias, cybersecurity—including spoofing attacks, lack of training data, opaque algorithms, algorithmic errors, and slow compute time may hinder the more advanced versions of fully autonomous systems. Moreover, all these challenges have salience in not just military projects, but also across the full spectrum of AI applications. Financial technology (FinTech), healthcare, autonomous vehicles—these and many more uses need basic research to ensure their safety and effectiveness.

Algorithmic Bias. Until we create code that can efficiently write new code itself, algorithms will be programmed by humans, who have favoritisms and prejudices. Moreover, AI training datasets are inherently finite and can reflect gender and racial bias. For example, the Microsoft facial recognition software Rekognition was used by the ACLU (American Civil Liberties Union) in a test on Congressmen against 25,000 criminal mug shots. Bizarrely, there were 28 matches of Congressmen to criminals with a seeming preference for minorities, implying that facial recognition (face-rec) software holds racial bias and inherent errors. Amazon claims that the ACLU used the software improperly, applying only the default 80% threshold for a match. [33] Before such software can be deployed in spaces in which people’s lives are at stake, it may be warranted to step back and execute more verification and validation (V&V) of its processing abilities, test for and remove built-in bias, and get a better handle on what are deemed acceptable thresholds for success. “A technology that’s proven to vary significantly across people based on the color of their skin is unacceptable in 21st-century policing." [34] Such proofs will be especially important when face-rec is deployed in autonomous weapons, which may be making life and death decisions based on its output.

Cybersecurity. No computer algorithm is immune to hacking. An example concern for AI is that of spoofing of images to make the AI think it is seeing an image different than reality. A recent article in Science clarified the dilemma that AI researchers now confront, and that weapons developers must deal with: “Impressive advances in AI—particularly machine learning algorithms that can recognize sounds or objects after digesting training data sets—have spurred the growth of living room assistants and autonomous cars. But these AIs are surprisingly vulnerable to being spoofed.” [35] A few stickers can change a stop sign into a speed limit sign. Adding a few pixels undetectable to a human eye can change a turtle into a rifle. Thus, even though there has been incredible progress made in image analysis via AI, with profound implications for autonomous weapons, it is presently still all too easy to spoof the image and confuse the algorithms. Both improved detection and countermeasures must be researched to diminish the fallibility of AI code.

Coding Errors. A challenge with any autonomy is the sheer complexity of the code. “Sophisticated automation requires software with millions of lines of code: 1.7 million for the F-22 fighter jet, 24 million for the F-35 jet, and some 100 million lines of code for a modern luxury automobile.” [ 5 ] No single human or even team of humans can catch all the bugs in code, since even if the software appears bug-free in beta testing, when it gets out into the real world unexpected circumstances can yield surprises. Obviously, this is undesirable in a system with weapons attached to it. What is needed is not just human-developed code, but also code that can fix its own code. Researchers such as at Carnegie Mellon University are developing code that can scan other code and fix problems automatically. [36] As systems become increasingly complex, such self-correcting code will become even more critical. Research into this area needs to increase, but with the caveat that we should also have safeguards and tracking systems in place to ensure traceability of the corrections inserted anew into the web of code. It is critical to make the autonomous systems both less brittle and more verifiable.

Black Box. Any military commander should understand why actions are taken, especially if they result in loss of life or valuable physical assets. Unfortunately, AI using machine learning is mostly a black box—once a result is produced, there is no way to assess how the result was obtained. A DARPA program, Explainable AI , seeks to change that: “Dramatic success in machine learning has led to a torrent of AI applications. Continued advances promise to produce autonomous systems that will perceive, learn, decide, and act on their own. However, the effectiveness of these systems is limited by the machine’s current inability to explain their decisions and actions to human users. The DoD is facing challenges that demand more intelligent, autonomous, and symbiotic systems. Explainable AI—especially explainable machine learning—will be essential if future warfighters are to understand, appropriately trust, and effectively manage an emerging generation of artificially intelligent machine partners.” [37]

Limited Training Data. A two year old child can see a ball once and then recognize other balls as play objects forever, but an AI using classic deep learning needs thousands of examples to recognize future objects as those of the same object class. Research is underway now to diminish this need for copious data, including [as noted from [38] ]:

· Capsule Networks , which record hierarchal networks to learn with less data

· Deep Reinforcement Learning , which combines reinforcement learning with deep neural networks to learn by interacting with the environment

· Generative Adversarial Networks , which use a type of unsupervised deep learning system, implemented as two competing neural networks, enabling machine learning with less human intervention

· Lean and Augmented Data Learning, which leverages different techniques—such as transfer learning or one-shot learning—that enable a model to learn from less data or synthetic data

· Hybrid Learning Models— which combine different types of deep neural networks with probabilistic approaches to model uncertainty

Ultimately, reducing the amount of training data needed is especially critical in autonomous systems, as there are challenges in obtaining real-world data. (One would not initiate a battle just to train an AI on battlefield conditions.)

Slow Compute. When operating an autonomous system, whether a civilian autonomous vehicle or a military drone, how fast calculations can be made is a critical decider of functionality. Especially for moving objects, calculations and decisions need to be made at millisecond or faster rates to avoid collisions. “Edge computing,” which enables on-device or system computing as opposed to sending and receiving signals to and from a central compute hub such as a cloud computing center, is a strong trend in AI. Of course, computing on-device requires faster, more mobile computer packages. NVIDIA and other companies are investing heavily in edge computing to bring the calculations onto the device, and thus short-circuit the need for signal delays. [39] With the on-going trends for diversification of computer chips beyond Moore’s Law, edge computing will need to be more researched and invested in to enable autonomous systems. [40] , [41]

A MIDDLE GROUND?

Many in the AI community understandably express concern about development and deployment of autonomous weapons. Nevertheless, despite numerous calls for banning autonomous weapons, several nation states with the military will and the economic means are moving forward to create such intelligent systems.

However, numerous technical challenges exist that may compromise the functionality and effectiveness of any autonomous weapons. The aforementioned AI issues must be overcome to minimize or completely avoid weapons that are in the least worst case dysfunctional and potentially harmful even to those deploying them, and in the absolute worst case possessing potential to go rogue on their human operators.

A key question every AI researcher must ask is whether we would prefer to be on the unprotected receiving end of countries that ignored bans and moved forward in developing autonomous weapons (that may be rife with algorithmic errors, bias and other issues outlined above), or would we rather be in a carefully planned defensive position ourselves with proper safeguards and other issues addressed. A growing chorus of op-eds advocate for engagement by AI researchers with the defense community because of these concerns. [42] , [43]

History has shown that nations that fell behind in technology advancement were subsequently outmatched on the battlefield. “In 1588, the mighty Spanish Armada was defeated by the British, who had more expertly exploited the revolutionary technology of the day: cannons. In the interwar period between World War I and World War II, Germany was more successful in capitalizing on innovations in aircraft, tanks, and radio technology and the result was the blitzkrieg—and the fall of France. The battlefield is an unforgiving environment. When new technologies upend old ways of fighting, militaries and nations don’t often get second chances to get it right.” [ 5 ]

Perhaps there is a middle ground in which proponents and opponents of autonomous weapons might meet. We propose that AI researchers execute basic research —in concert with both civilian and defense agencies—for the benefit of all autonomous systems efforts. Without engaging in the actual construction of weapons, AI researchers could support the US Government in its efforts to overcome the technical issues listed above. For example, countermeasures to autonomous weapons cannot be developed without a detailed understanding of the functionality of those weapons. Many US federal agencies—DARPA, IARPA, NSF, etc. [44] —offer funding that is targeted to basic science.

Such research would benefit the wider economy, also. FinTech, health technology, autonomous vehicles, energy systems, law enforcement—these and many more AI sectors would benefit greatly from research into the topics above.

Thus, instead of carte-blanche refusals to work with the US Government on AI issues that might lead to weapons development, perhaps a middle ground exists to execute basic research and thus aid both industry and defense.

CONCLUSIONS

History will most probably repeat itself—calls for banning autonomous weapons will be overwhelmed by already significant investments, concerns over adversary actions, and potential for battlefield dominance. Well-meaning people will sign letters, resign their employment, or otherwise protest the development and use of autonomous weapons. International organizations will hold conferences and sign more comprehensive treaties. But when push comes to shove and nation states see their adversaries tooling-up or applying weaponry powered by AI and robotics, they will have little hesitation to go all-in on autonomous systems. The signs are already there with the United States, China and Russia initiating large programs. Proliferation to other countries is only a matter of time, as we have seen with other technologies (3d-printing, smartphones, etc.)

There is an inexorable push toward autonomous weapons being developed and deployed by nation states with the will and economic backing to do so. The AI community must remain diligent, and, in keeping with its own personal or organizational ethics, should consider engaging in basic science research to benefit both civilian and military applications as the technology advances.

NOTES (all links accessed August 2018)

[1] Hague Convention, https://www.britannica.com/event/Hague-Conventions

[2] A.Q. Arbuckle, “1914-1918, Balloons of World War I, https://mashable.com/2016/03/02/wwi-balloons/#tSQb8HZBe8qo

[3] Unrestricted Submarine Warfare, https://en.wikipedia.org/wiki/Unrestricted_submarine_warfare

[4] N. Eardley, “Files show confusion over Lusitania sinking account,” 1 May 2014, https://www.bbc.co.uk/news/uk-27218532

[5] Paul Scharre, “Army of None: Autonomous Weapons and the Future of War,” 2018, W. Norton & Company, Inc.

[6] “Allied submarines in the Pacific War,” https://en.wikipedia.org/wiki/Allied_submarines_in_the_Pacific_War

[7] C. Pellerin, “Deputy Secretary: Third Offset Strategy Bolsters America’s Military Deterrence”, 31 OCT 2016, https://www.defense.gov/News/Article/Article/991434/deputy-secretary-third-offset-strategy-bolsters-americas-military-deterrence/

[8] M. Rosenberg, J. Markoff, “The Pentagon’s ‘Terminator Conundrum’: Robots That Could Kill on Their Own,” 25 OCT 2016, https://www.nytimes.com/2016/10/26/us/pentagon-artificial-intelligence-terminator.html

[9] “Background on Lethal Autonomous Weapons Systems in the CCW,” https://www.unog.ch/80256EE600585943/(httpPages)/8FA3C2562A60FF81C1257CE600393DF6?OpenDocument

[10] K. Johnson, “Google CEO bans autonomous weapons in new AI guidelines,” 7 JUNE 2018, https://venturebeat.com/2018/06/07/google-ceo-bans-autonomous-weapons-in-new-ai-guidelines/

[11] M. McFarland, “Leading AI researchers vow to not develop autonomous weapons,” 18 JUL 2018, https://money.cnn.com/2018/07/18/technology/ai-autonomous-weapons/index.html

[12] “Lethal Autonomous Weapons Pledge,” https://futureoflife.org/lethal-autonomous-weapons-pledge/?cn-reloaded=1

[13] C. Pash, “These tech leaders have signed a pledge against killer robots,” 19 JUL 2018, https://www.weforum.org/agenda/2018/07/the-worlds-tech-leaders-and-scientists-have-signed-a-pledge-against-autonomous-killer-robots

[14] G.M. Del Prado, “Stephen Hawking, Elon Musk, Steve Wozniak and over 1,000 AI researchers co-signed an open letter to ban killer robots,” 27 JUL 2015, https://www.businessinsider.com/stephen-hawking-elon-musk-sign-open-letter-to-ban-killer-robots-2015-7

[15] A. Mehta, “DoD stands up its artificial intelligence hub,” 29 JUN 2018, https://www.c4isrnet.com/it-networks/2018/06/29/dod-stands-up-its-artificial-intelligence-hub/

[16] J. Doubleday, “DOD wants $75 million to establish Joint AI Center, forecasts $1.7B over six years,” 18 JUL 2018, https://insidedefense.com/daily-news/dod-wants-75-million-establish-joint-ai-center-forecasts-17b-over-six-years

[17] Command, Control, Communications, Computers, Intelligence, Surveillance and Reconnaissance

[18] “AlphaGo versus Lee Sedol,” https://en.wikipedia.org/wiki/AlphaGo_versus_Lee_Sedol

[19] R. Faruk, “Military Implications of AlphaGo,” 30 MAY 2017, http://actionablethought.com/military-implications-of-alphago/

[20] E. Kania, “China's Artificial Intelligence Revolution,” 27 JUL 2017, https://thediplomat.com/2017/07/chinas-artificial-intelligence-revolution/

[21] V. Barhat, “China is determined to steal A.I. crown from US and nothing, not even a trade war, will stop it,” 4 MAY 2018, https://www.cnbc.com/2018/05/04/china-aims-to-steal-us-a-i-crown-and-not-even-trade-war-will-stop-it.html

[22] T. Campbell, “Kardashev Scale Analogy: Long-Term Thinking about Artificial Intelligence,” 5 NOV 2017, https://www.futuregrasp.com/kardashev-scale-analogy

[23] Unmanned Aerial Vehicles

[24] S.N. Romaniuk and T. Burgers, “China’s Swarms of Smart Drones Have Enormous Military Potential,” 3 FEB 2018, https://thediplomat.com/2018/02/chinas-swarms-of-smart-drones-have-enormous-military-potential/

[25] Unmanned underwater vehicles

[26] People’s Liberation Army

[27] P. Glass, “China’s Robot Subs Will Lean Heavily on AI: Report,” 23 JUL 2018, https://www.defenseone.com/technology/2018/07/chinas-robot-subs-will-lean-heavily-ai-report/149959/?oref=d-mostread

[28] A. Conn, “Podcast: Six Experts Explain the Killer Robots Debate,” 31 JUL 2018, https://futureoflife.org/2018/07/31/podcast-six-experts-explain-the-killer-robots-debate/

[29] J. Vincent, “Putin says the nation that leads in AI ‘will be the ruler of the world’,” 4 SEP 2017, https://www.theverge.com/2017/9/4/16251226/russia-ai-putin-rule-the-world

[30] “Putin: Leader in artificial intelligence will rule world,” 4 SEP 2017, https://www.cnbc.com/2017/09/04/putin-leader-in-artificial-intelligence-will-rule-world.html

[31] P. Tucker, “Russia to the United Nations: Don’t Try to Stop Us From Building Killer Robots,” 21 NOV 2017, https://www.defenseone.com/technology/2017/11/russia-united-nations-dont-try-stop-us-building-killer-robots/142734/

[32] S. Bendett, “Here’s How the Russian Military Is Organizing to Develop AI,” 20 JUL 2018, https://www.defenseone.com/ideas/2018/07/russian-militarys-ai-development-roadmap/149900/

[33] S. Hollister, “Amazon facial recognition mistakenly confused 28 Congressmen with known criminals,” 26 JUL 2018, https://www.cnet.com/news/amazon-facial-recognition-thinks-28-congressmen-look-like-known-criminals-at-default-settings/

[34] B. Barrett, “Lawmakers Can't Ignore Facial Recognition's Bias Anymore,” 26 JUL 2018, https://www.wired.com/story/amazon-facial-recognition-congress-bias-law-enforcement/

[35] M. Hutson, “Hackers easily fool artificial intelligence,” Science , 20 July 2018, 361 (6399), pp. 215.

[36] Research at SCS, Carnegie Mellon University, https://www.cs.cmu.edu/research

[37] D. Gunning, “Explainable Artificial Intelligence (XAI),” DARPA, https://www.darpa.mil/program/explainable-artificial-intelligence

[38] D.S. Ganguly, “Top Artificial Intelligence trends in 2018-You Should Consider For AI & Machine Learning Development,” 3 JUL 2018, https://hackernoon.com/top-trends-in-2018-you-need-to-follow-for-artificial-intelligence-machine-learning-development-dfeee0eabb5f

[39] J. Markman, “This Is Why You Need To Learn About Edge Computing,” 3 APR 2018, https://www.forbes.com/sites/jonmarkman/2018/04/03/this-is-why-you-need-to-learn-about-edge-computing/#2b04f1571a56

[40] T. Campbell, R. Meagley, “Next-Generation Compute Architectures Enabling Artificial Intelligence - Part I of II,” 2 FEB 2018, https://www.futuregrasp.com/next-generation-compute-architectures-enabling-artificial-intelligence-part-I-of-II

[41] T. Campbell, R. Meagley, “Next-Generation Compute Architectures Enabling Artificial Intelligence—Part II of II,” 8 FEB 2018, https://www.futuregrasp.com/next-generation-compute-architectures-enabling-artificial-intelligence-part-II-of-II

[42] G.C. Allen, “AI researchers should help with some military work,” 6 JUN 2018, https://www.nature.com/articles/d41586-018-05364-x

[43] P. Luckey, T. Stephens, “Silicon Valley should stop ostracizing the military,” 8 AUG 2018, https://www.washingtonpost.com/opinions/silicon-valley-should-stop-ostracizing-the-military/2018/08/08/7a7e0658-974f-11e8-80e1-00e80e1fdf43_story.html?utm_term=.674b3489c79d

[44] Defense Advanced Research Project Agency, Intelligence Advanced Research Project Activity, National Science Foundation

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Opportunities and Challenges from Artificial Intelligence for Law Enforcement

Fri, 20 Jul 2018 17:07:10 GMT

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Law enforcement (LE) has unique challenges—identification of individuals, prediction of criminal actions, tracking of money flows, tagging and defending against fake news, and enhanced communications across agencies, among others—that demand rapid data acquisition and analysis. As artificial intelligence (AI) offers unprecedented capabilities to acquire and analyze big data, it is no surprise that the LE community is interested in applying AI for its purposes. [1] But given LE’s unique requirements on speed, low error rates, and accurate predictions, is AI up to the task yet of helping LE?

We offer here a review of several facets of AI that could benefit or compromise LE. [2] As AI becomes more advanced, the potential for its application by LE experts (and criminals) will only increase. Let us hope that LE officers in the field and analysts in the office are able to leverage AI to protect people around the world.

DATA, DATA, DATA—THE INTERNET OF THINGS

We are living in a veritable information explosion. Every day we create roughly 2.5 quintillion bytes of new data. [3] The Internet of Things – i.e., the universe of devices connected to the internet – is forecasted to exceed 30 billion devices by 2030. Over a trillion sensors of all forms is projected to be in use by 2020. [4] This digital tsunami tasks our local and federal governments in gathering data, comprehending it, and crafting sound policies. Because of the rapid-response needs of LE, this information must be acquired, curated and analyzed in almost real-time.

AI can assist in managing this surfeit of data. Especially well-positioned here is deep learning (DL), which requires massive datasets in order to yield accurate output. Thus, leveraging DL makes perfect sense for the LE community; each of the technologies described below use some aspect of DL.

FACIAL RECOGNITION (FACE-REC)

There are some people who never forget a face. So-called ‘super-recognizers,’ such individuals can remember people they haven’t seen in decades. For the rest of us, we can use mnemonic tricks to not be embarrassed in conversations. Theoretically, computers should trump everyone on identifying people, as fundamentally face-rec is a challenge of data recall.

In the early days of face-rec competitions with algorithms, we routinely beat our silicon creations…then DL was developed. During an imaging contest using a large database, the Imagenet , one team in 2012 successfully leveraged large computational power with DL to improve face-rec accuracy dramatically. As clarified by MIT Technology Review:

“This was the first time that a deep convolutional neural network had won the competition, and it was a clear victory. In 2010, the winning entry had an error rate of 28.2 percent, in 2011 the error rate had dropped to 25.8 percent. But SuperVision won with an error rate of only 16.4 percent in 2012 (the second best entry had an error rate of 26.2 percent). That clear victory ensured that this approach has been widely copied since then.” [5]

At this writing, the record for face-rec is claimed to be less than one percent error—although exact error values vary as a function of lighting, amount of the face exposed, motion and other factors. Depending on the humans tested, errors by people attempting to recognize others range the gamut from 0% to 50% error. [6]

Recognizing faces is a critical aspect of policing work, yet the techniques for its execution must be trusted. Mug shots and line-ups have been used for decades with human victims attempting to identify criminals. Nevertheless, errors can occur and people can be falsely jailed based upon a mistaken identity. [6] Using algorithms to reduce errors can help, but we must be careful to not be too reliant on them, as even a small error rate can still yield false arrests. [7] Microsoft recently called for the US Government to consider regulating face-rec: “Facial recognition technology raises issues that go to the heart of fundamental human rights protections like privacy and freedom of expression. These issues heighten responsibility for tech companies that create these products. In our view, they also call for thoughtful government regulation and for the development of norms around acceptable uses. In a democratic republic, there is no substitute for decision making by our elected representatives regarding the issues that require the balancing of public safety with the essence of our democratic freedoms. Facial recognition will require the public and private sectors alike to step up – and to act.” [8]

Law enforcement should proceed with caution on the use of face-rec. At the present stage of its development, algorithmically-generated face-rec should be verified independently by humans to avoid mistaken identities.

CAMERAS

Cameras of all forms (including body cameras-bodycams, CCTV-closed circuit television, and drone cameras) can be a powerful tool for reducing violence both at and from police, as well as identifying criminals. The range of intrusiveness of these systems varies from passive (observation only) to active (seeking matches to faces and other crime scene clues in large databases).

On the passive side, bodycams can be used to protect both the citizen and the police officer, but they have their own limitations. “They record audio and video, often with date and time stamps as well as GPS coordinates. They can also be Bluetooth-enabled and set to stream in real time. Some have to be turned on manually, others can be triggered automatically by, for instance, an officer unholstering his weapon.” [9] Bodycams are controversial, and their effectiveness in reducing violence both at and from police officers has been called into question. A 2017 study, claiming to be the largest on bodycams to date, concluded that these systems offer mixed results. “A new study suggests police body cameras have had no measurable effect in Washington, D.C., when it comes to civilian complaints, use of force incidents, policing activity or judicial outcomes involving the city’s Metropolitan Police Department.” [10] Nevertheless, we should note that these are still early days for video recordings of citizen-police interactions, and that more use and data may offer greater clarity on the effectiveness of this LE tool.

On the active side, machine learning and deep learning can be leveraged to identify a wide range of characteristics of people and objects from cameras. Perhaps the most intrusive application of AI for personal identification via cameras is in China:

“It should give Westerners no comfort that China—a one-party state obsessed with social order—is at the forefront of developing bodycams. One Beijing company says it has invented a shoulder-worn, networked model that can recognize faces. Another Chinese firm has equipped police with facial-recognition cameras embedded in glasses that are meant to let officers know in real time if they are looking at someone on a police blacklist. One estimate values China’s surveillance-tech market in 2018 at $120bn. Human-rights campaigners fear that such technology has already been used to monitor activists, enabling arbitrary detention.” [ 7]

Although privacy limitations may limit such active scanning and analysis in Western countries, the genie is out of the bottle for many nation states. Cameras with AI built-in for face-rec are now commercially available. [11] , [12] It remains to be seen how far LE takes such tools. Applications of cameras have strong implications for other technologies discussed below, however.

UNMANNED AERIAL VEHICLES (UAVs, aka, DRONES)

UAVs have been commercially available for over a decade, but their price-points and capabilities came to levels only just within the past few years so that LE is adopting them. UAVs range in size from an artificial bee [13] to a large aircraft [14] . Their capabilities are just as diverse, with not just cameras, but also weaponization [15] , fire-fighting [16] , and even tree-planting [17] . Only a lack of imagination limits what one can do with these flying machines.

Privacy advocates have been calling for years to place limitations on drone use. After much debate, the US Federal Aviation Administration (FAA, https://www.faa.gov/ ) put into laws the conditions under which one must operate to fly a drone of any kind in public. For example, UAVs are not allowed to fly in US national parks without a permit (for concerns of stressing wildlife, endangering people, etc.). [18] [19] All drones within certain size range can be flown legally only with a permit. [20]

LE may have looser restrictions on UAV applications – for example, with a known threat of terrorism or a manhunt underway for a suspected felon. Nevertheless, even LE must comply in the United States with certain limitations to protect the public’s privacy and safety. “Local and state agencies in the US have passed rules imposing strict regulations on how law enforcement and other government agencies can use drones. Indiana law specifies that police departments can use drones for search-and-rescue efforts, to record crash scenes and to help in emergencies, but otherwise a warrant is required to use a drone. That means police probably won’t be able to fly them near large gatherings unless a terrorist attack or crime is under way.” [21]

The implications of UAVs are extensive for LE. The ability to have a birds-eye view practically anywhere (both outdoors and indoors) offers significant upside for LE in tracking, identification and analysis of criminal individuals and actions. Nevertheless, as noted above, UAV use also brings with it great responsibility to ensure safety and privacy of the public. As UAVs advance technologically, LE must continue to engage with local and national governments to ensure these issues are jointly addressed. INTERPOL is taking a lead in assessing drone threats and challenges to LE via its Global Complex for Innovation. [22]

FAKE NEWS

Fake news has a long history. Ever since printed text became widely available, fake news has been used by people to push propaganda and influence one’s enemies. “Fake news took off at the same time that news began to circulate widely, after Johannes Gutenberg invented the printing press in 1439. ‘Real’ news was hard to verify in that era. There were plenty of news sources—from official publications by political and religious authorities, to eyewitness accounts from sailors and merchants—but no concept of journalistic ethics or objectivity. Readers in search of fact had to pay close attention. In the 16th century, those who wanted real news believed that leaked secret government reports were reliable sources, such as Venetian government correspondence, known as relazioni . But it wasn’t long before leaked original documents were soon followed by fake relazioni leaks. By the 17th century, historians began to play a role in verifying the news by publishing their sources as verifiable footnotes. The trial over Galileo’s findings in 1610 also created a desire for scientifically verifiable news and helped create influential scholarly news sources.” [23]

Since those early days, fake news has expanded to include every new recording or communications technology--see image at bottom of text. When photography was first introduced to the public, faking a photograph taken on real film required staging shots or chemical manipulation. Digital photos make it much easier to simply jigger a few pixels and produce any photo one wishes. Now it is a challenge to have a photo be admitted as evidence in a court of law.

As the internet came online in the latter part of the 20th century, fake news really took off. Now with major online platforms such as Facebook and Twitter, as well as the 24/7 news cycle, we struggle to identify what is real or not while living in our chosen filter bubbles. Nation states such as Russia and China have embraced social media as a means for pushing their own messages or propaganda.

We are now progressing into even more sophisticated fake news. Audio and video can be manipulated at will. One research group claims that all they need is 3.7 seconds of a human’s voice soundwaves to mimic them with an algorithm. [24] Another research group takes such work to another level by superimposing a researcher’s face onto that of another individual, and consequently enabling the researcher to make the victim say anything with any expression the researcher desires. [25]

Further into the realm of difficulty and complexity, augmented and virtual realities (AR and VR) will become the next-gen platforms for fake news. As AR and VR are used more by companies and governments, we may fully expect them to be target platforms for fake news. Essentially, any mode of communications is now fair game for fake news generation.

Law enforcement has an important role to play in fake news, albeit one that appears to be still gelling in terms of its scope. Police officers and nation states must both identify and defend against fake news. During the recent Mexico elections (July 2018), an effort was made to identify fake news in text, audio and video by AI. Human curators then reviewed the tagged media and decided if it was truly false and what to do with it if so. [26] At this writing, it is unclear how effective this work was.

As we expand how we exchange news, nefarious individuals will continue to manipulate our communications and distribute their own narratives or influence efforts. It is upon the shoulders of LE and nation states to ensure that fake news is identified and does not result in skewed elections or criminal actions. Sadly, the veracity of news is a competition between criminals and LE with no foreseeable end.

PREDICTIVE POLICING

In the 2002 movie Minority Report , the character played by Tom Cruise uses a bio-based computing system to predict crimes before they happen and thus arrest perpetrators before they commit their nefarious deeds. We are a ways away from being able to make such fully accurate predictions—and accurate they must be; imagine being falsely accused for murder because of algorithmic error or bias?—but we may still consider that the summation of the technologies described above and others are leading us down this path. Whether and how quickly predictive policing moves out of fiction into every police station remains to be seen. Nevertheless, significant efforts and research dollars are being expended to make predictive policing become the norm in identifying crime locations and individuals in advance. We list below just a few examples of the research now underway, and already deployed tools, to enable future Tom Cruises to wield their predictive hands against crime.

In the United States, the company Palantir has been testing predictive policing in large cities for several years. Unbeknownst to several city council members, Palantir algorithms were used in New Orleans for over six years; the Palantir contract has since been cancelled. [27] The Los Angeles Police Department (LAPD) still uses Palantir’s models, but there are claims that the results have inherent racial bias. [28] Because of concerns of opaqueness (lack of disclosure of the databases) and bias, there are now major lawsuits pending against Palantir in Chicago, Los Angeles, New Orleans and New York City. [29]

Japan recently announced a major effort to enable predictive policing. “The AI-based system would employ a ‘deep learning’ algorithm that allows the computer to teach itself by analyzing big data. It would encompass the fields of criminology, mathematics and statistics while gathering data on times, places, weather and geographical conditions as well as other aspects of crimes and accidents.” [30] The Japanese government hopes to prove the system in at least one prefecture first, and then deploy it further at the 2020 Tokyo Olympics. [31]

Perhaps China has taken the level of predictive policing to its most intrusive level, violating international human rights norms in the process. In the prefecture of Xinjiang, local authorities and police use big data analytics to monitor a wide variety of personally identifiable information (PII) on minority citizens, often without their knowledge. Individuals can then be arrested and imprisoned without trial. [32] Chinese citizens, although said by Chinese officials to be trusting of the technology wielded by their local law enforcement, are known to cover their faces when cameras are pointed at them. [33]

Thus, other than a few nation states such as China, widespread use of predictive policing is not yet the norm. An informal survey made during the recent INTERPOL—UNICRI meeting in Singapore on AI, robotics and law enforcement showed that none of the 20 attending countries had significant predictive policing programs planned or in place at this writing (China was not present at the meeting). [ 1] This lack of broad application may change, though, as algorithm, data and computational capabilities increase. Ultimately, caution should be exercised by all parties in deployment of predictive policing to ensure that it does not compromise the privacy of innocent civilians and/or result in ill will and lawsuits against LE that may take the technology too far.

CONCLUSIONS & IMPLICATIONS

One can argue that AI is certainly capable of helping to address some of the data acquisition and analysis needs in LE, but with the caveat that its use must be monitored and proofed before arrests or litigation decisions are made. Technologies such as face-rec with algorithm-enabled digital cameras, as well as nascent experiments in predictive policing, offer LE officers new tools by which to perform their tasks. The challenges of fake news are increasing in complexity and may require AI to assist LE in its identification and digital removal. Nevertheless, inherent limitations of AI such as inherent bias, opaque algorithms, and lack of complete datasets should give us pause before unchecked deployment of AI in LE. In the United States and elsewhere there are major lawsuits pending against LE agencies that may have deployed AI algorithms too soon without rigorous verification and validation (V&V) mechanisms in place. Other nations such as China have gone all-in on AI-enabled LE capabilities, but at the cost of human rights for their citizens who in minor cases lose privacy and in worst cases get falsely arrested and jailed without warrants.

Despite technical and ethical challenges, the use of AI by LE is entering an inexorable state. In many criminal cases, there is simply too much data for the traditional officer or detective alone to capture and to assess all relevant evidence. Whether and how individual police departments, small towns, cities and nation states apply the powers of AI algorithms remains to be seen. Moreover, its adoption will most probably be driven by culture and nation state interests. Let us hope that the applications of AI in LE are made after careful consideration of both the benefits to LE agencies and the citizens they are charged to protect.

NOTES (all websites accessed July 2018)

[1] Witness the recent workshop, “1st INTERPOL-UNICRI Global Meeting on the Opportunities and Risks of Artificial Intelligence and Robotics for Law Enforcement 2018,” held in Singapore on 11-12 July 2018. I presented the closing talk, “Looking Into The Future…”

[2] Note this review is not meant to be comprehensive, but merely an introduction to the field. Substantial literature has been written about each of the overview sections here.

[3] “10 Key Marketing Trends for 2017,” IBM Marketing Cloud, https://www-01.ibm.com/common/ssi/cgi-bin/ssialias?htmlfid=WRL12345USEN

[4] “Over 1 trillion sensors could be deployed by 2020,” January 26, 2017, https://www.electronicspecifier.com/around-the-industry/over-1-trillion-sensors-could-be-deployed-by-2020

[5] “The Revolutionary Technique That Quietly Changed Machine Vision Forever,” September 9, 2014, https://www.technologyreview.com/s/530561/the-revolutionary-technique-that-quietly-changed-machine-vision-forever/

[6] “Combining the facial recognition decisions of humans and computers can prevent costly mistakes,” June 4, 2018, https://theconversation.com/combining-the-facial-recognition-decisions-of-humans-and-computers-can-prevent-costly-mistakes-97365

[7] “CEO of a Facial Recognition Company: The Tech is Too Volatile to Give to Law Enforcement,” June 28, 2018, https://www.nextgov.com/emerging-tech/2018/06/ceo-facial-recognition-company-tech-too-volatile-give-law-enforcement/149337/

[8] “Facial recognition technology: The need for public regulation and corporate responsibility,” Microsoft Blogs, June 13, 2018, https://blogs.microsoft.com/on-the-issues/2018/07/13/facial-recognition-technology-the-need-for-public-regulation-and-corporate-responsibility/?ranMID=24542&ranEAID=je6NUbpObpQ&ranSiteID=je6NUbpObpQ-kig_5rwdCle73fiVClXDjg&epi=je6NUbpObpQ-kig_5rwdCle73fiVClXDjg&irgwc=1&OCID=AID681541_aff_7593_1243925&tduid=(ir_W8kVW1XEs3ns3NhV1m05%3ATJDUkjVTBXmPxzmUg0)(7593)(1243925)(je6NUbpObpQ-kig_5rwdCle73fiVClXDjg)()&irclickid=W8kVW1XEs3ns3NhV1m05%3ATJDUkjVTBXmPxzmUg0

[9] “Data Detectives,” Technology Quarterly, Justice, The Economist, June 2, 2018.

[10] “New Study Casts Doubt On Effectiveness Of Police Body Cameras. But Is That Fair?,” October 24, 2017, https://www.huffingtonpost.com/entry/dc-police-body-camera-study_us_59ee1bace4b003385ac11440

[11] “Artificial intelligence software suite enables advanced people detection capabilities,” June 25, 2018, https://www.vision-systems.com/articles/2018/06/artificial-intelligence-software-suite-enables-advanced-people-detection-capabilities.html

[12] “AWS DeepLens - Deep learning enabled video camera for developers,” https://www.amazon.com/dp/B075Y3CK37/x_tw_01_01_01_04

[13] “Rise Of The Robot Bees: Tiny Drones Turned Into Artificial Pollinators,” March 23, 2017, https://www.npr.org/sections/thesalt/2017/03/03/517785082/rise-of-the-robot-bees-tiny-drones-turned-into-artificial-pollinators

[14] “Predator RQ-1 / MQ-1 / MQ-9 Reaper UAV,” https://www.airforce-technology.com/projects/predator-uav/

[15] “Homeland Security: Warns of Weaponized Drones as Terror Threat,” November 10, 2017, https://www.usnews.com/news/national-news/articles/2017-11-10/homeland-security-warns-of-weaponized-drones-as-terror-threat

[16] “How firefighters are using drones to save lives,” August 27, 2017, https://www.cnbc.com/2017/08/26/skyfire-consulting-trains-firefighters-to-use-drones-to-save-lives.html

[17] “New tree-planting drones can plant 100,000 trees in a single day,” January 31, 2018, https://www.techspot.com/news/73032-new-tree-planting-drones-can-plant-100000-trees.html

[18] Years ago, I was in Yosemite National Park and saw a small UAV being used to film a valley, despite a large crowd present in the same area. I asked the operator if he had a permit—the individual feigned surprise, then immediately landed the drone, grabbed it, and took-off in his pickup truck.

[19] “Unmanned Aircraft in the National Parks,” https://www.nps.gov/articles/unmanned-aircraft-in-the-national-parks.htm

[20] FAA Drone Zone, https://faadronezone.faa.gov/#/

[21] “What do drones, AI and proactive policing have in common? Data,” https://www.sas.com/en_us/insights/articles/risk-fraud/drones-ai-proactive-policing.html#.WzAdhrg3kD4.linkedin

[22] “Drones: The New Threat and Challenge to Law Enforcement: Threat-Tool-Evidence,” INTERPOL, https://www.INTERPOL.int/

[23] “The Long and Brutal History of Fake News,” December 18, 2016, https://www.politico.com/magazine/story/2016/12/fake-news-history-long-violent-214535

[24] “‘Deep Voice’ Software Can Clone Anyone's Voice With Just 3.7 Seconds of Audio,” March 7, 2018, https://motherboard.vice.com/en_us/article/3k7mgn/baidu-deep-voice-software-can-clone-anyones-voice-with-just-37-seconds-of-audio

[25] “New Digital Face Manipulation Means You Can’t Trust Video Anymore,” May 13, 2016, https://singularityhub.com/2016/05/13/new-digital-face-manipulation-means-you-cant-trust-video-anymore/#sm.0006vb0tkhrkef311um15lzmcfb3c

[26] “AI to help tackle fake news in Mexican election,” June 30, 2018, https://www.bbc.com/news/technology-44655770

[27] “New Orleans ends its Palantir predictive policing program,” March 15, 2018, https://www.theverge.com/2018/3/15/17126174/new-orleans-palantir-predictive-policing-program-end

[28] “The LAPD Has A New Surveillance Formula, Powered By Palantir,” May 8, 2018, https://theappeal.org/the-lapd-has-a-new-surveillance-formula-powered-by-palantir-1e277a95762a/

[29] “'Predictive policing': Big-city departments face lawsuits,” July 5, 2018, http://www.chicagotribune.com/news/sns-bc-us--predictive-policing-challenges-20180705-story.html

[30] “Japan considers crime prediction system using big data and AI,” June 24, 2018, https://www.japantimes.co.jp/news/2018/06/24/national/crime-legal/japan-mulling-crime-prediction-using-big-data-ai/#.WzEJPKdKg2y

[31] “Japan trials AI-assisted predictive policing before 2020 Tokyo Olympics,” January 29, 2018, https://www.scmp.com/news/asia/east-asia/article/2130980/japan-trials-ai-assisted-predictive-policing-2020-tokyo-olympics

[32] “China: Big Data Fuels Crackdown in Minority Region,” February 26, 2018, https://www.hrw.org/news/2018/02/26/china-big-data-fuels-crackdown-minority-region

[33] “Looking Through the Eyes of China’s Surveillance State,” July 16, 2018, https://www.nytimes.com/2018/07/16/technology/china-surveillance-state.html?partner=IFTTT

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Why Stop at Voice?

Tue, 19 Jun 2018 15:46:29 GMT

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Sitting in a coffee shop this morning, I heard a man at the table next to me take a call on his smartphone. He proceeded to speak so loudly that all the patrons had no difficulty hearing all the nuances of his small business. It was quite distracting, so I departed the comfortable environment earlier than I’d planned. Incidents like this make me wonder about the future of voice as the dominant control method for computers in the future. Voice is all the buzz (no pun intended) in Silicon Valley. Many technology leaders (Apple, Amazon, Microsoft, etc.) are betting that verbal commands will soon be the major means of communication for our numerous devices. But is verbal chatter the terminal point for our computer communications? Are we destined solely for the Star Trek Enterprise’s command prompts, to be stating the contemporary equivalent, even in public forums, of “Computer, take me to the Bridge”?

TODAY’S VOICE TECH

We naturally try to emulate our biological capabilities in the computer. Why wouldn’t we want to just talk to our devices and have them both accurately record our words, as well as chat with us like a human would? As a result of intense interest in deploying voice commands, tremendous advances have been made since the early days of voice recognition. Algorithms powered by artificial intelligence (AI) now can achieve up to 95 percent accuracy in understanding voices, on par with humans. [1] Depending how fast one can type vs. speak, voice commands offer the potential to be more efficient, and perhaps more accurate, than the output from our standard QWERTY keyboards. [2]

Top tech firms are embracing voice—Amazon has Alexa, Apple has Siri, and Microsoft has Cortana. We can talk to small receiving stations like Amazon’s Echo and Google’s Home and order just about anything…all by using our natural voice. Companies such as IPsoft have even captured a real human’s likeness and digitized it into an efficient enterprise software system to answer repeatedly, and to millions of customers simultaneously, mundane questions about insurance and human resource related matters. [3] Despite these advances and promises, there exist challenges with voice that may compromise its full deployment.

CHALLENGES WITH VOICE

Even we humans frequently misinterpret what each other is saying, so it’s no surprise that algorithms also make mistakes. Several particular linguistic challenges come to mind that could be onerous for a machine to overcome completely:

Noisy environments. The “Cocktail Party problem” is encountered frequently at conference receptions. All of us have a unique voice, even if it can change a little based upon our mood, fatigue or alcohol consumption. Nevertheless, throw in a bunch of other voices and background noise, and we can be challenged in interpreting the voice that we wish to focus upon, even if the speaker is right next to us. And that’s with a lifetime of learning; for an algorithm with little such experience, such a situation can be even more difficult. To cut through the cacophony of extraneous noise, one approach by Google is to incorporate both audio and visual cues using a machine learning algorithm trained upon thousands of YouTube videos. [4] Further advances may be needed to ensure no misinterpretation of a speaker in noisy environments.

Confusing command prompts. A recent story highlights the dangers of being completely trusting of a verbal command system. A family with an Amazon Echo system accidentally had an entire private dialogue of theirs emailed to a colleague several miles away – without the family’s knowledge. Per Amazon:

“Echo woke up due to a word in background conversation sounding like ‘Alexa’. Then, the subsequent conversation was heard as a ‘send message’ request. At which point, Alexa said out loud ‘To whom?’ At which point, the background conversation was interpreted as a name in the customer’s contact list. Alexa then asked out loud, ‘ contact name , right?’ Alexa then interpreted background conversation as ‘right’.” [5]

Better safeguards – for example, double-prompts, or additional codes and keywords – may be needed for such systems to ensure conversations meant to remain private are kept private.

Accents. Accents are one of the joys of language, but they also can be a burden for interpretation by an algorithm. Impressively, Amazon Alexa, Google Home, and Apple Siri test surprisingly well against speakers with multiple accents saying identical phrases, but there are still occasional failures. [6] Interpreting different accents may be compounded by other factors such as background noise.

Slang and dialects. Two beautiful aspects of language are slang words and dialects that can be almost entirely distinct from a mother tongue. For example, in German language ‘Umgangssprache’ (slang) can be an intense experience for even the fluent speaker. Throw in the differences among ‘Hochdeutsch’ (high German), ‘Plattdeutsch’ (low German), ‘Schweizerdeutsch’ (Swiss German), and other dialects even from village to nearby village, and one can become confused quickly. Prior to my post-doctoral studies in Germany, my wife and I spent four months intensively studying Hochdeutsch and felt ourselves reasonably fluent enough to go into stores and order items with no issues. Then we moved to a village 10 kilometers outside the city where we learned German and went to a local grocery store. It seemed they had a different word or pronunciation for almost every item that we had ordered easily back in the big city. The issue was a dialect, almost distinct from Hochdeutsch, that we had not been taught. Voice algorithms working in such environments would have to learn all such nuances to be proficient.

Language changing. Another great thing about languages is that they are never static. New words are constantly being introduced, especially in areas of technology. Words also change their root meaning, progressing through so-called ‘soft changes’: “Words tend to pick up different usages and even meanings over time, often very remarkably. The word terrific used to have a highly negative meaning—something that terrifies. Only recently has it become a positive term.” [7] Just as humans must adapt to the morphing of a language, algorithms will also have to constantly maintain updated dictionaries and be on the lookout for words that are used in different ways than commonly heard.

Foreign languages. There are almost 7,000 globally recognized languages, although one might consider that among this Tower of Babel, we really only use about 10 or 20 languages prominently around the world. [8] , [9] Roughly two billion people speak only three languages (Chinese--including 10 varieties, Spanish and English). [10] Nevertheless, to truly become the main, go-to, approach for computer commands, must our algorithms become the equivalent of the Star Wars C3PO, whose fictional robotic abilities could accommodate over six million galactic tongues? Would introducing more languages also compound interpretation among them all? For example, could a word that was pronounced by a Chinese Mandarin speaker be misinterpreted, and thus mistranslated, as Korean? When Amazon Alexa was debuted in France, Amazon had to consider the occasional English American word creeping into French conversation. More data is critical to understand the nuances of Romance languages with multiple versions of “you” and other honorifics. [11]

BEYOND VOICE

Aside from writing or typing, we have many non-verbal ways of communicating – facial expressions, eye motions, musculoskeletal movements, even our thoughts themselves can be harnessed by a computer to convey what we mean. Brain-computer interfaces (BCI) have been researched for decades in attempts to tease out our thoughts. Similar to audio commands, however, challenges with all of them exist. The ability of a computer algorithm to interpret our thoughts is a research avenue that has attracted substantial interest, especially in the last few years. With the goal of discussing technologies that could be more widely accepted in the near term, we describe below exclusively examples of ex-vivo (out of the body) capabilities. [12]

Facial expressions can be used to detect basic wishes. A Brazilian team is working to use facial expressions to control a wheelchair for individuals lacking other means of controlling mobility. “The camera can identify more than 70 facial points around the mouth, nose and eyes. By moving these points, it is possible to get simple commands, such as forward, backward, left or right and, most importantly, stop.” [13]

Tracking pupil oscillations (pupillometry) is another approach for non-verbal commands. Researchers in France and The Netherlands showed in 2016 a new BCI method via pupillometry. “In our method, participants covertly attend to one of several letters with oscillating brightness. Pupil size reflects the brightness of the selected letter, which allows us–with high accuracy and in real time–to determine which letter the participant intends to select. The performance of our method is comparable to the best covert-attention brain-computer interfaces to date, and has several advantages: no movement other than pupil-size change is required; no physical contact is required (i.e. no electrodes); it is easy to use; and it is reliable. Potential applications include: communication with totally locked-in patients, training of sustained attention, and ultra-secure password input.” [14]

Measurement of thoughts directly is also generating strong interest in the research community. In 2016, Elon Musk announced the creation of the company Neuralink , which seeks to create an external system to read the electrical signals from one’s mind and thus control computers and merge with artificial intelligence (AI). [15] “Neuralink is developing ultra-high bandwidth brain-machine interfaces to connect humans and computers.” The venture raised over $27 million in late 2017. [16] A wonderful posting in Wait But Why goes into detail on the underlying technology. [17]

In April 2018, a graduate student at MIT created a system that uses electrical signals from muscles to capture thoughts without speaking. Called AlterEgo, the system was tested on 10 subjects and achieved 92 percent accuracy. “AlterEgo is a closed-loop, non-invasive, wearable system that allows humans to converse in high-bandwidth natural language with machines, artificial intelligence assistants, services, and ‘movements—simply by vocalizing internally.” Potential applications include air traffic control and military communications. [18]

IMPLICATIONS & CONCLUSIONS

Living in the mountains of Colorado, I rarely hear much traffic, planes or other noises that are common in more populous urban areas. But every time I travel to a big city, I am struck by the almost constant cacophony of sounds. Noise pollution is a recognized health issue. [19] While talking to a computer may not be a major contributor to noise, every decibel can be a distraction, whether it be in a coffee shop, in a cube farm at work, or at home with your spouse and kids. Non-verbal communication approaches have the potential to create a quieter world.

Individuals who have lost the ability to speak or communicate otherwise certainly could leverage non-audio BCI systems. Paraplegics, autistics and other handicapped people (for example, those with a speech impediment) could benefit tremendously from a system that is well-trained to their thoughts. Being trapped in one’s mind with no means of communication must be torture for such disease or accident victims. BCIs have great potential to help them engage more with the world.

Perhaps in the future we may extend the use of BCIs into the science fiction realm of telepathy, also. Imagine being able to communicate non-verbally with your relatives or friends, not only while in the vicinity of each other, but perhaps many miles apart. We already do this now with texting, but using our thumbs still requires physical contact with a smartphone (and can be very dangerous when driving). Using a BCI such as those described above could enable more seamless communications in many life situations.

Ultimately, instead of having to answer a cell phone call verbally in a coffee shop, might we soon instead be able to communicate effectively without saying a word? The technologies to get us there are in labs and test computer algorithms now. It remains to be seen how soon such capabilities will be widely available. Sometime soon, we might instead think , “Computer, order me a triple latte.”

NOTES (all websites accessed June 19, 2018)

[1] K. Wiggers, “Qualcomm claims its on-device voice recognition is 95% accurate,” May 25, 2018, https://venturebeat.com/2018/05/25/qualcomm-claims-its-on-device-voice-recognition-is-95-accurate/

[2] QWERTY keyboards were initially designed to actually slow us down in typing so that the manual typewriter keys would not jam. It’s funny how legacy technologies remain in modern use even if they are sub-optimal.

[3] S. Kesler, “Inside the bizarre human job of being a face for artificial intelligence,” June 5, 2017, https://qz.com/996906/inside-the-bizarre-human-job-of-being-a-face-for-artificial-intelligence/

[4] L. Tung, “Google AI can pick out a single speaker in a crowd: Expect to see it in tons of products,” April 13, 2018, https://www.zdnet.com/article/google-ai-can-pick-out-a-single-speaker-in-a-crowd-expect-to-see-it-in-tons-of-products/

[5] S. Wolfson, “Amazon's Alexa recorded private conversation and sent it to random contact,” May 24, 2018, https://www.theguardian.com/technology/2018/may/24/amazon-alexa-recorded-conversation

[6] M. Calore, “Watch People With Accents Confuse the Hell Out of AI Assistants,” May 16, 2017, https://www.wired.com/2017/05/ai-assistants-accented-english/

[7] A. Myers-Stanford, “How artificial intelligence can teach itself slang,” June 7, 2017, https://www.futurity.org/deep-learning-language-1452822-2/

[8] S.R. Anderson, “How many languages are there in the world?,” 2010, Linguistic Society of America, https://www.linguisticsociety.org/content/how-many-languages-are-there-world

[9] “List of languages by number of native speakers,” https://en.wikipedia.org/wiki/List_of_languages_by_number_of_native_speakers

[10] J. Myers, “These are the world’s most spoken languages,” February 22, 2018, https://www.weforum.org/agenda/2018/02/chart-of-the-day-these-are-the-world-s-most-spoken-languages

[11] B. Barrett, “Inside Amazon's Painstaking Pursuit to Teach Alexa French,” June 13, 2018, https://www.wired.com/story/how-amazon-taught-alexa-to-speak-french/

[12] These examples are not intended to be an exhaustive review of the BCI field, but merely demonstrative of the art of the possible.

[13] A. Pasolini, “Wheelchair controlled by facial expressions to hit the market within 2 years,” May 19, 2016, https://newatlas.com/wheelchair-facial-commands/43206/

[14] Sebastiaan Mathôt, Jean-Baptiste Melmi, Lotje van der Linden, and Stefan Van der Stigchel, “The Mind-Writing Pupil: A Human-Computer Interface Based on Decoding of Covert Attention through Pupillometry,” PLoS One. 2016; 11(2): e0148805. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4743834/

[15] T. Lacoma, “Everything you need to know about Neuralink: Elon Musk’s brainy new venture,” November 7, 2017, https://www.digitaltrends.com/cool-tech/neuralink-elon-musk/

[16] Neuralink, Crunchbase, https://www.crunchbase.com/organization/neuralink

[17] T. Urban, “Neuralink and the Brain’s Magical Future,” April 20, 2017, https://waitbutwhy.com/2017/04/neuralink.html

[18] AlterEgo, MIT Media Lab, https://www.media.mit.edu/projects/alterego/overview/

[19] N. Lee, D. Anderson, J. Orwig, “Noise pollution is a bigger threat to your health than you may think, and Americans aren't taking it seriously,” January 26, 2018, http://www.businessinsider.com/noise-pollution-effects-human-hearing-health-quality-of-life-2018-1

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Quantum Supremacy is Near

Fri, 04 May 2018 15:10:40 GMT

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Toward Quantum Supremacy [1]

Can you imagine being able to solve complex problems almost instantaneously? What would normally take months would take minutes. Seemingly impossible questions would become straightforward to answer. In the world of quantum computing, we are close to doing just that. Quantum supremacy, the point at which a quantum computer exceeds the capabilities of any classical computer [2] , is now near at hand. “A successful demonstration of quantum supremacy would prove that engineered quantum systems, although still in their infancy, can outperform the most advanced classical computers.” [3]

For over 50 years, Moore’s Law—with the international support of the industrial consortium SEMATECH—drove the central processing unit (CPU) to achieve a doubling of the number transistors roughly every two years. The trillion-dollar semiconductor industry was reliant upon this exponential; it arguably enabled the entire information technology (IT) industry that we have today. Recently, though, Moore’s Law has faltered and the classic transistor number doubling within two years no longer occurs. Clock speeds have also slowed since the integrated circuits (ICs) simply fry and fail if they are run too fast. Moreover, the cost of a new fabrication (fab) plant for ICs now exceeds $10 billion, meaning only the largest companies can afford them. Essentially, the semiconductor industry can no longer maintain Moore’s Law with traditional silicon-based ICs.

But the death of Moore’s Law does not mean that advanced computing ceases its acceleration. [4] On the contrary, technologies such as QC are picking up where CPUs falter. Blazing fast computational speed is the leverage that the nonlinear QC architecture offers beyond the linear computing CPU. QC has the potential to spot patterns extremely quickly within large data sets. “It might be possible for the quantum computers to access all items in your database at the same time to identify similarities in seconds.” [5]

Traditional semiconductor chips operate with bits that are either ‘on’ or ‘off’ depending on their electrical state; quantum computers work differently and have unique technical challenges. Quantum computing (QC) uses qubits , in which an individual bit can be in one of three states: on, off, and, uniquely, both on and off simultaneously. Moreover, QC demands cooling of the processing chip to fractions of a single digit Kelvin to avoid disturbing processing in the qubits. QC also requires wholly new approaches to software since the binary on/off paradigm is no more, and instructions do not load sequentially, but rather execute simultaneously .

But how rapidly is QC advancing? Harkening back to Moore’s Law, I was curious if QC might be following a similar development path. Taking all the available literature data [6] , I graphed the number of qubits available from multiple research efforts over time. The criteria for a data point’s inclusion are that the qubit result has demonstrated controlled, coherent manipulation of individual quantum objects, such as multi-qubit gates or generation of mutual qubit entanglement. This strict criteria obviates inclusion at this writing of a number of recent, higher number qubit announcements, such as IBM’s and Intel’s independent 49 qubit chipsets [7] , [8] and Google’s 72 qubit chipset [9] .

Amazingly, the result is another exponential. With a residual-squared graph fit of ~0.91 (decent for a random data plot), it is shown that the number of qubits doubles roughly every 7.3 years.

How many qubits are needed to achieve quantum supremacy is up for debate, but certainly a higher number is recognized as necessary. The threshold number will be driven by factors such as the demonstrated application of multi-qubit gates or the generation of mutual qubit entanglement, as well low error rates for a given qubit chipset. A low error rate might allow quantum supremacy to be achieved at around 50 qubits. “For quantum supremacy, Google will need to build a 49-qubit system with a two-qubit fidelity [a measure of how error-prone the process is] of at least 99.7 percent [where a higher number denotes lower error].” [10] Thus, assuming the graph fit stays the same in coming years, quantum supremacy might be achieved by roughly 2026, eight years from now. However, Google believes that quantum supremacy might be achieved within even 2018 [ 9] , assuming their 72 qubit chipset possesses low error rates, and meets our graph’s criteria.

Whatever the final qubit number necessary, we are arguably looking at achieving quantum supremacy within the next few years at most.

The implications of quantum supremacy will be profound. “…what is really exciting about quantum computing is that we have good reason to believe that a quantum computer would be able to efficiently simulate any process that occurs in Nature.” [11] Potential applications of QC range over the whole economic spectrum. As detailed on the website of Rigetti Computing [12] , a Silicon Valley startup that recently met the graph’s criteria for 19 qubits, QC can be applied in difficult problems in areas such as physics, chemistry, materials science, medicine, transportation and logistics, and neuroscience and artificial intelligence (AI). The challenge for applying quantum supremacy will be not so much finding problems that need tackling, but rather crafting the problems in ways that a quantum computer can handle them.

Challenges Remaining

Of course, technical and business challenges remain before any research group or company can claim the trophy of quantum supremacy; I note three categories:

Hardware. The pursuit of more qubits for a given multi-qubit chipset or collection of entangled qubits is of paramount importance. We should expect to see more announcements in the coming months and years of increasing qubit numbers. Assuming we maintain an exponential graph fit as shown in the figure, organizations developing quantum computers will certainly be exciting to watch.

Programming. Programming of QC remains a challenge…how do you code for a computer that uploads and executes all its instructions simultaneously? QC programming, as we noted in the BootstrapLabs Applied AI Conference panel on the “Death of Moore’s Law,” [ 4] presents a tremendous growth opportunity for startups. Companies such as Rigetti Computing already offer access through an API (application program interface) for QC in the cloud. [13] Programmers interested in exploring the opportunities presented by QC are encouraged to engage with such QC leaders.

Scale-up. Like any new technology, scale-up will be an issue for QC. Just because something works well in the lab, doesn’t mean it will work at scale when thousands of programmers wish to access a quantum computer that has demonstrated quantum supremacy. More investment and greater recognition within the computer science community of the power of QC will help ensure wider QC availability.

…not Conclusions

It is challenging to make hard conclusions on QC today, because this incredibly promising field is really just starting. Companies both big (for example, Google, IBM, Intel) and smaller (for example, Rigetti Computing) are investing hundreds of millions of dollars in the chase for quantum supremacy and the consequent ability to leverage the power of the quantum world. Industry recognizes that QC is now approaching a technical inflection point. The barrier at which problems that cannot be solved by any classical computer might soon be overcome with the quantum computer.

Acknowledgements

I gratefully acknowledge Mr. Madhav Thattai of Rigetti Computing for insights and select data points on the figure, as well as his comments on our panel at the BootstrapLabs Applied AI Conference in San Francisco, CA on April 12, 2018.

Notes

[1] Note that an alternative term preferred by some researchers is “quantum advantage.” The distinction between supremacy and advantage is straightforward: supremacy is a benchmark, advantage is an application. [personal communication, Madhav Thattai, Rigetti Computing]

[2] J. Preskill, 13 NOV 2012, “Quantum computing and the entanglement frontier,” WSPC – Proceedings, arXiv:1203.5813v3, https://arxiv.org/pdf/1203.5813.pdf

[3] Neill, et al., 13 APR 2018, “A blueprint for demonstrating quantum supremacy with superconducting qubits,” Science, 360 (6385), 195-198.

[4] T. Campbell, 12 APR 2018, “Death of Moore’s Law—and the Birth of New Semiconductor Industries and AI,” BootstrapLabs Applied AI Conference, San Francisco, CA, keynote and panel.

[5] B. Marr, 5 SEP 2017, “How Quantum Computers Will Revolutionize Artificial Intelligence, Machine Learning And Big Data,” Forbes, https://www.forbes.com/sites/bernardmarr/2017/09/05/how-quantum-computers-will-revolutionize-artificial-intelligence-machine-learning-and-big-data/#2d0bf8df5609

[6] Graph modified from “Moore’s Law for Quantum Computers,” 2011, http://www.quantenblog.net/physics/moores-law-quantum-computer ; links for all data point sources are listed below:

Source	Year	Qubits
https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.75.4714	1995	2
https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.82.1345	1999	3
https://www.nature.com/articles/35005011	2000	4
https://www.nature.com/articles/nature02643	2004	5
https://www.nature.com/articles/nature04279	2005	8
https://www.nature.com/articles/nphys1603	2010	10
https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.106.130506	2011	14
https://www.technologyreview.com/s/544421/googles-quantum-dream-machine/	2015	9
https://medium.com/rigetti/unsupervised-machine-learning-on-rigetti-19q-with-forest-1-2-39021339699	2017	19
https://techcrunch.com/2017/11/10/ibm-passes-major-milestone-with-20-and-50-qubit-quantum-computers-as-a-service/	2017	20

[7] E. Pednault, 27 OCT 2017, “Quantum Computing: Breaking Through the 49 Qubit Simulation Barrier,” IBM Blog, https://www.ibm.com/blogs/research/2017/10/quantum-computing-barrier/

[8] J. Hsu, 9 JAN 2018, “CES 2018: Intel's 49-Qubit Chip Shoots for Quantum Supremacy,” IEEE Spectrum, https://spectrum.ieee.org/tech-talk/computing/hardware/intels-49qubit-chip-aims-for-quantum-supremacy

[9] J. Kelly, 5 MAR 2018, “A Preview of Bristlecone, Google’s New Quantum Processor,” https://research.googleblog.com/2018/03/a-preview-of-bristlecone-googles-new.html .

[10] M. Reynolds, 22 JUN 2017, “Google on track for quantum computer breakthrough by end of 2017,” https://www.newscientist.com/article/2138373-google-on-track-for-quantum-computer-breakthrough-by-end-of-2017/

[11] J. Preskill, 27 JAN 2018, “Quantum computing in the NISQ era and beyond,” arXIV:1801.00862v2

[12] https://rigetti.com/

[13] https://www.rigetti.com/forest

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Next-Generation Compute Architectures Enabling Artificial Intelligence—Part II of II

Thu, 08 Feb 2018 13:06:01 GMT

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

In this second of two installments (for Part I, see https://www.futuregrasp.com/next-generation-compute-architectures-enabling-artificial-intelligence-part-I-of-II ), we examine technologies emerging to enable the next generations of commercial AI. Across these exciting approaches, common trends noted in the first wave of AI entrants in the marketplace continue. These include:

· strategies to reduce noise and increase efficiency

· smaller scale engineering

· a proliferation of advanced materials

· increasing speed, density and efficiency

· diversification in mechanism

Nanomaterials

Nanomaterials are considered promising for extending Moore’s Law beyond CMOS [1] as fundamental transistors, as well as copper interconnects connecting the transistors, to enable a whole new class of integrated circuits (ICs). Carbon based nanomaterials, such as carbon nanotubes (CNTs) and graphene, are especially interesting because of their rapid transport of electrons. [2] CNTs can be made in various diameters and chiralities [3] to optimize electron transport for a given application, and may be grown in situ to facilitate fabrication. Graphene can be vapor deposited as a layer and shaped with lithography. [4] Both materials show desirable switching and optical characteristics, as well as good properties for electronic and optical architectures ( vide infra ). [5] On the horizon are materials with even more favorable transport and switching properties, such as MoS2. [6]

Applying nanomaterials such as CNTs can enable whole new chip architectures. Traditional CPUs are two-dimensional (2D) - ICs are produced with numerous layers of valleys, bridges and interconnects on a flat silicon substrate. Multiple ICs are connected on the 2D landscape of a circuit board. Unfortunately, by being 2D the processing speed is inherently limited in CPUs because of the longer time that electrons must take to transport between ICs.

ICs, and thus AI, would benefit with a more 3D architecture. Recent work by MIT and Stanford demonstrates the promise of 3D architectures built on a backbone of millions of CNTs. “As applications analyze increasingly massive volumes of data, the limited rate at which data can be moved between different chips is creating a critical communication ‘bottleneck.’ And with limited real estate on the chip, there is not enough room to place them side-by-side, even as they have been miniaturized.” [7] [8] On-chip storage and processing speed are both increased by sandwiching CNTs between existing silicon chips. This novel approach offers advantages over traditional IC design. “In perspective this chip is important because of its efficiency (low power consumption thanks to the carbon nanotube and fast processing thanks to the colocation of storage and processing) which makes it very interesting in AI applications.” [9]

Of course, a major challenge with incorporating nanomaterials into traditional IC production is to minimize the changes to semiconductor fabs. Hundreds of billions of dollars have been invested over the last 50 years to optimize IC manufacturing with a silicon architecture. Any new material will have to fit within this process flow at minimal cost. At the end of the day, semiconductor production is all about economics - getting the best and most chips manufactured at the lowest cost.

Optical

Another emerging semiconductor technology is optical computing. Instead of electrons, light (photons) is used as the medium to conduct information between transistors. Thus, computations can literally occur at the speed of light, faster than conventional electron-driven chips. Materials enabling optical computing are in general quite different from electronic materials, and materials for optics that can mesh with ASIC and other nanoelectronic fabrication strategies are subject to intensive research. Silicon (and silicon-germanium) have a strong lead for integrated photonic chips; however, its limitations are driving advances in graphene and other materials. [10] [11] [12]

As with any advanced computational architecture, faster processing speed can benefit AI computation. Deep learning especially requires rapid data analysis with large datasets, all while saving energy compared to traditional CPUs or GPUs. Optical chips hold the potential to offload the matrix calculations needed for deep learning to another chip from the CPU, and thereby save power consumption and minimize space use by multiple CPUs. [13] Recently, collaborators from the University of Bristol , Microsoft, Google, Imperial College , Max Planck Institute , and the Sun Yat-sen University invented a new algorithm for solving the energy structure of quantum systems on optical quantum computers. Their device provides an example of quantum algorithm development implemented on an optical device architecture that can yield mass manufacturable systems. [14]

Groups such as at MIT [15] [16] , the University of Arizona [17] , and the University of Alberta in Canada [18] are driving research in advanced optical materials, switches, lasers and nano-optics to further optical computing. For example, a group in the UK recently shrunk the distance a given set of photons must travel on a chip from more than a centimeter to mere micrometers, thus enabling even faster processing speeds. [19] However promising optical chips might be, much research remains for optical computing to enable more calculations on a given chip than nanoelectronics, and as well to increase scalability for full systems capable of AI. However, advances in this space have been advancing in line with other architectures, and we may expect to see greater deployment of light on chips in the future.

DNA and Other Biological Compute Strategies

Just as neuromorphic is a higher level of neural net integration, adapted from observations of biological computing systems, it is no surprise that it is possible to further emulate and co-opt biochemical information processing systems. DNA [20] is the stuff of life, so wouldn’t it be ironic if we were to design a computer that enables AI with the materials from our own genomes? That approach is one of the most exotic in our list of advanced computer architectures that may someday create highly adaptable parallel processing, and high memory storage density ICs.

“DNA computing is a form of parallel computing in that it takes advantage of the many different molecules of DNA to try many different possibilities at once.” [21] [22] One recent claim by the University of Manchester is for a DNA computer that ‘grows’ as it computes, i.e., its compute pathways (the DNA) copy and replicate themselves as a computation is underway to find a given answer faster. [23] This is wholly different from a fixed silicon IC design.

Another major advantage of DNA is in its potential for memory storage. “Capable of storing 215 petabytes (215 million gigabytes) in a single gram of DNA, the system could, in principle, store every bit of datum ever recorded by humans in a container about the size and weight of a couple of pickup trucks.” [24] Microsoft announced in early 2017 that they plan on adding DNA storage to their cloud. [25] While promising, a barrier for such deployment is cost – DNA storage requires both read and write capabilities, which at this writing cost several thousand dollars per megabyte—too expensive for large-scale cloud storage. But as costs come down and read/write speeds and accuracy improve, we may have some of our future social media posts being stored on the very material we depend upon in our cells.

Other molecules have also been adapted for biological compute. [26] RNA and proteins have been used in biochemical/biological computing architectures. [27] Both of these biomolecules interact strongly with each other and their environments to a greater degree than DNA. As they also have programmable structures, a language of RNA and protein shapes that fit like puzzle pieces and interact like gear teeth is ubiquitous for biological information processing. These shapes can form switches and logic gates, and enable massively parallel processing via chemistry held on surfaces and in liquid solution. [28]

One limitation to biological compute (BC) currently, is that, like quantum computing (QC), BC has challenges in manipulating algorithms and in structuring problems that early limited systems can solve. BC has demonstrated real advantages in certain kinds of complex optimization problems also attractive for QC. [29] [30] A few intrepid startups are already active in this space; for example, Rebel Bio recently demonstrated a DNA-based biological computer that plays Tetris. [31] With increased investment in this area, it is anticipated that systems will continue to improve in generality, flexibility, and convenience. [32]

Integration of BC from bacterial systems has been well established. [33] Other species including slime molds harnessed to solve complex problems and integrate environmental sensing. [34] [35] [36] Far beyond simple organisms, ambitious progress has been made interfacing the nanoelectronic approaches with the biological, culminating in the development of AI implanted in a prosthetic to aid human cognition, as being commercialized in a partnership among INTENT LTD , Intel, and Qualcomm . [37] While BC is perhaps the most forward-leaning AI embodiment, it presents a unique challenge as it is the origin of human consciousness. [38]

Conclusions

We emphasized in Part I that “There is now no longer a single Moore’s Law, but rather a suite of them, with each new law a function of compute architecture, software and application.” The additional emerging technologies discussed here in Part II further embrace and extend the portent of this observation. [39]

Diversification in device and integration of devices offer technology options that enable niche applications. As these new capabilities are proven successful, some niches will expand into more general applications. [40] These developments set the stage for waves of M&A [41] as smaller firms are absorbed into larger concerns producing higher volume.

Ultimately, new compute architectures will ensure many additional generations of continuous improvement, well beyond Moore’s Law. AI will thus be driven to ever higher levels of compute and more capable performance.

Notes

[1] Complementary metal-oxide semiconductor

[2] Z. Chen, et al., “Carbon nanotubes for high-performance logic,” MRS Bulletin, 39, August 2014, 719-726.

[3] CNTs come in multiple sizes, including single-walled and multi-walled, as well as chiralities, or ‘handedness,’ such as zigzag and armchair. Such diverse CNT species are tailored toward the desired electrical and other properties. [ https://en.wikipedia.org/wiki/Carbon_nanotube , accessed January 2018.]

[4] https://www.intechopen.com/books/graphene-materials-advanced-applications/fundamentals-of-chemical-vapor-deposited-graphene-and-emerging-applications , accessed January 2018.

[5] http://aip.scitation.org/doi/abs/10.1063/1.4795332?journalCode=apl , accessed January 2018.

[6] https://www.sciencedirect.com/science/article/pii/S1369702115003776 , accessed January 2018.

[7] http://news.mit.edu/2017/new-3-d-chip-combines-computing-and-data-storage-0705 , accessed January 2018.

[8] http://sites.ieee.org/futuredirections/2017/07/10/carbon-nanotubes-powering-the-chip-of-the-future/ , accessed January 2018.

[9] Ibid.

[10] http://iopscience.iop.org/article/10.1088/2040-8978/18/7/073003/meta , accessed January 2018.

[11] http://ieeexplore.ieee.org/document/7409809/ , accessed January 2018.

[12] https://web.stanford.edu/group/heinz/publications/Pub212.pdf , accessed January 2018.

[13] https://www.scientificamerican.com/article/light-powered-computers-brighten-ai-rsquo-s-future/ , accessed January 2018.

[14] http://advances.sciencemag.org/content/4/1/eaap9646 , accessed January 2018.

[15] http://newsoffice.mit.edu/2013/computing-with-light-0704 , accessed January 2018.

[16] https://www.nature.com/articles/nphoton.2017.93 , accessed January 2018.

[17] http://www.ece.arizona.edu/ua-leads-multi-university-team-optical-computing-research , accessed January 2018.

[18] http://www.wirelessdesignmag.com/news/2014/08/electrical-engineers-take-major-step-toward-photonic-circuits?et_cid=4108958&et_rid=281046610&location=top , accessed January 2018.

[19] https://www.sciencedaily.com/releases/2017/11/171130141048.htm , accessed January 2018.

[20] Deoxyribonucleic acid

[21] https://en.wikipedia.org/wiki/DNA_computing#cite_note-16 , accessed January 2018.

[22] Lewin, D. I. (2002). "DNA computing". Computing in Science & Engineering. 4 (3): 5–8. doi : 10.1109/5992.998634 , accessed January 2018..

[23] http://rsif.royalsocietypublishing.org/content/14/128/20160990 , accessed January 2018.

[24] http://www.sciencemag.org/news/2017/03/dna-could-store-all-worlds-data-one-room , accessed January 2018.

[25] https://www.technologyreview.com/s/607880/microsoft-has-a-plan-to-add-dna-data-storage-to-its-cloud/ , accessed January 2018.

[26] https://books.google.com/books?hl=en&lr=&id=fKNJvbRwrY8C&oi=fnd&pg=PA179&ots=_VrR2ZCW-k&sig=wjTl7gnFvFhov8-7uvuvLoYP-oY#v=onepage&q&f=false , accessed January 2018.

[27] https://biodesign.asu.edu/news/living-computers-rna-circuits-transform-cells-nanodevices , accessed January 2018.

[28] http://www.cbc.ca/news/mcgill-bio-super-computer-1.3466095 ?, accessed January 2018.

[29] https://link.springer.com/article/10.3103/S0146411617050108 , accessed January 2018.

[30] https://www.lunduniversity.lu.se/article/using-nanotechnology-to-create-parallel-computers , accessed January 2018.

[31] https://rebelbio.co/fancy-playing-tetris-bio-computer/ , accessed January 2018.

[32] https://blogs.wsj.com/venturecapital/2015/11/18/andreessen-horowitz-launches-200m-bio-computer-fund-hires-ex-stanford-prof/ , accessed January 2018.

[33] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5264507/ , accessed January 2018.

[34] https://www.sciencedaily.com/releases/2014/03/140327100335.htm , accessed January 2018.

[35] https://phys.org/news/2011-09-biological-cancer-cells.html , accessed January 2018.

[36] https://www.technologyreview.com/s/512821/shrinking-blob-computes-traveling-salesman-solutions/ , accessed January 2018.

[37] https://www.nds.ox.ac.uk/news/researchers-take-major-step-forward-in-artificial-intelligence , accessed January 2018.

[38] http://ubiquity.acm.org/article.cfm?id=1103047 , accessed January 2018.

[39] In 2015, one author [TAC] ran a workshop titled “Moore’s Law 2.0.” One conclusion from the series of presentations by industry and academia is that we now have such a diverse realm of compute architectures that there is confusion in both industry and the US Government on which approach to fund the most. No one knows which chipset will be the winner, if there even will be a clear winner.

[40] https://www.cbinsights.com/research/top-acquirers-ai-startups-ma-timeline/ , accessed January 2018.

[41] Mergers & Acquisitions

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

Next-Generation Compute Architectures Enabling Artificial Intelligence - Part I of II

Fri, 02 Feb 2018 16:56:03 GMT

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration

In this first of two installments, we explore the emergence and impact of compute architectures behind artificial intelligence (AI) now in the marketplace. In Part II (see https://www.futuregrasp.com/next-generation-compute-architectures-enabling-artificial-intelligence-p... ) we look at emerging compute technologies that might influence the next generations of commercial AI.

Death of Moore’s Law

In 1965, Dr. Gordon E. Moore wrote an article based on a trend he noticed: the number of transistors in an integrated circuit (IC) doubles approximately every two years [1] . Fueled by unrelenting demands from more complex software, faster games and broadband video, Moore’s Law has held true for over 50 years. [2] It became the de-facto roadmap against which the semiconductor industry drove its R&D and chip production, as facilitated by SEMATECH’s pre-competitive agreements among global semiconductor manufacturers. Recently, that roadmap has faltered due to physics limitations and the high cost-benefit economics incurred by the incredibly small scales that chip manufacturing has reached. Electron leakages and difficulties shaping matter at the single-digit nanometer scales of the transistors fundamentally limit further miniaturization. So many electrons are being moved though such tight spaces so quickly that there is an entire field in the semiconductor industry devoted just to chip cooling; without thermal controls, the ICs simply fry and fail. A new fabrication plant (fab) can cost more than $10 billion, severely limiting the number of companies able to produce denser ICs.

Despite the looming end of Moore’s Law, computationally-intensive artificial intelligence (AI) has exploded in capabilities in the last few years – but how, if compute is slowing down? The solution to exceeding compute limitations of traditional von Neumann style central processing units (CPUs) [3] has been to invent and leverage wholly new architectures not dependent on such linear designs [4] .

A veritable zoo of compute architectures – including GPUs [5] , ASICs [6] , FPGAs [7] , quantum computers, neuromorphic chips, nanomaterial-based chips, optical-based ICs, and even biochemical architectures - are being researched and/or implemented to better enable deep learning and other instantiations of AI. Here we review the latest / greatest [8] of non-CPU computer architectures relevant to AI. In each section, we describe the hardware, its impact to AI, and a selection of companies and teams active in its R&D and commercialization.

GPU

Graphics processing units (GPUs) have taken the computing world by storm in the last few years. Fundamentally, GPUs operate differently than CPUs. “A simple way to understand the difference between a GPU and a CPU is to compare how they process tasks. A CPU consists of a few cores optimized for sequential serial processing, while a GPU has a massively parallel architecture consisting of thousands of smaller, more efficient cores designed for handling multiple tasks simultaneously.” [9]

As deep learning requires high levels of computational power, AI is a particularly strong application for GPUs. Neural network implementation on GPU-based systems accelerates the training process dramatically compared to traditional CPU architecture; this saves substantial time and money. [10]

Nvidia is one of the earliest companies to apply GPUs to AI; its latest TITAN V PC GPU is designed specifically for deep learning in a desktop platform. [11] A US-based company, Nvidia is now a powerhouse that in Q3, 2017 controlled 72.8% of the discrete GPU market share. [12] As more OEMs [13] invest in R&D, we can expect continuous improvements in the rapidly advancing space of GPUs.

ASIC

An application-specific integrated circuit (ASIC) is an IC customized for a particular use, rather than intended for general-purpose use. [14] The advantage for AI is that an ASIC can be designed explicitly for fast processing speed and/or massive data analysis via deep learning. While ASIC processors typically implement traditional von Neumann architecture, foundry-produced ASIC chips with neural net architecture have found niche applications commercially.

There have been several press releases about companies going their own route with ASICs to become independent of general GPUs. [15] Realizing their loss of market share and wishing to not be wholly dependent upon Nvidia, numerous software firms and OEMs are designing and/or implementing their own ASICs. [16] Google is already on its second version of its TPU (tensor processing unit), designed to accelerate processing speeds in their vast server farms. [17] Intel has been working on its own AI-dedicated chip for several months. [18] Intel claims its Nervana Neural Network Processor family, or NNP for short, is on track to increase deep learning processing speeds by 100 times by 2020. [19] Tesla recently announced its efforts to build from scratch its own AI chips to have better control and faster processing speed in its electric cars. [20]

FPGA

A field-programmable gate array (FPGA) enables greater customization after manufacturing—ergo, field-programmable . “The FPGA configuration is generally specified using a hardware description language (HDL), similar to that used for an application-specific integrated circuit (ASIC). FPGAs contain an array of programmable logic blocks, and a hierarchy of reconfigurable interconnects that allow the blocks to be ‘wired together’, like many logic gates that can be inter-wired in different configurations.” [21]

While somewhat larger in scale than analogous ASIC processors, a key advantage offered by FPGAs for AI is in their application of deep neural networks requiring analysis of large amounts of data. The flexibility of programming circuitry rather than instructions enables complex neural nets to be configured and reconfigured seamlessly in a standardizable platform. [22] Intel applies FPGAs toward such deep data problems. [23] [24] Their FPGAs were shown to be competitive to GPUs earlier this year:

“While AI and DNN research favors using GPUs, we found that there is a perfect fit between the application domain and Intel’s next generation FPGA architecture. We looked at upcoming FPGA technology advances, the rapid pace of innovation in DNN [deep neural network] algorithms, and considered whether future high-performance FPGAs will outperform GPUs for next-generation DNNs. Our research found that FPGA performs very well in DNN research and can be applicable in research areas such as AI, big data or machine learning which requires analyzing large amounts of data. The tested Intel Stratix 10 FPGA outperforms the GPU when using pruned or compact data types versus full 32 bit floating point data (FP32). In addition to performance, FPGAs are powerful because they are adaptable and make it easy to implement changes by reusing an existing chip which lets a team go from an idea to prototype in six months—versus 18 months to build an ASIC.” [25] [26]

Beyond the traditional semiconductor OEMs, numerous software companies are developing FPGAs to enable faster processing speeds within their data centers. In August 2017, Amazon , Baidu and Microsoft announced manufacturing of FPGAs targeted toward machine learning. “Taken together, these initiatives point to increased opportunities for other large AI consumers and providers to have their cake and eat it too; they can optimize custom chips for their applications while avoiding the cost and potential technology obsolescence of going down the custom ASIC approach.” [27]

Quantum

Traditional computers operate with bits that are either ‘on’ or ‘off’ depending on their electrical state; quantum computers work differently. Quantum computers use qubits , in which an individual bit can be in one of three states: on, off, and, uniquely, both on and off simultaneously. As with other advanced computational architectures, quantum computing (QC) is still in early stages of development. QC has unique hardware issues – for example, the need to cool the processing chip to fractions of a single digit Kelvin to avoid disturbing processing in the qubits. QC also requires a wholly new approach to software since the binary on/off paradigm is no more, and instructions do not load sequentially, but rather execute simultaneously , from which the speed originates. Potential areas where QC could have significant impact include energy (by accelerating discovery in high energy plasma manipulation, materials chemistry, and other optimization modeling problems), medicine (by modeling complex biochemical systems to accelerate drug discovery and understanding of disease), information (by enabling big data analytics and new encryption approaches), aerospace (by modeling hypersonic aerodynamics), and materials (by simulating complex materials systems).

Speed is the leverage that QC offers AI, especially in machine learning. QC has the potential to spot patterns extremely quickly within large data sets. “It might be possible for the quantum computers to access all items in your database at the same time to identify similarities in seconds.” [28] For an AI that must analyze billions of images or data points, to execute this analysis in seconds using the power of the simultaneous on-off qubit is huge compared to a sequential CPU that might take weeks or months for the same action.

Critical challenges to all QC strategies include reducing noise and maintaining system coherence. Improved materials and architectures will continue to drive QC development. [29]

The Canadian company D-Wave Systems is one of the leaders in QC. Although there is some controversy on whether D-Wave actually makes quantum computers (their systems are marketed as more limited quantum annealers [30] [31] ), their multi-million dollar systems have been purchased by Google , Lockheed Martin and others in the past few years. D-Wave recently installed their latest 2000Q System at NASA Ames . [32] Because it is so different than traditional computing, a core challenge with QC is the dearth of software dedicated to operating in both qubit states (on and off) simultaneously. D-Wave partnered with 1QBit to meet this challenge; for example, initial solutions of previously intractable protein folding problems are being explored by 1QBit using D-Wave’s systems.

Several other companies are active in hardware and/or software development for QC. IBM announced recently they reached a threshold of 50 qubits in operation; such a system would be the largest quantum computer to date. However, “…as revolutionary as this development is, IBM's 50-qubit machine is still far from a universal quantum computer.” [33] Microsoft also recently released a preview version of its Q# (pronounced “Q sharp”) software dedicated to quantum computing. “Microsoft's hope is that this selection of tools, along with the training material and documentation, will open up quantum computing to more than just physicists.” [34] Microsoft and IBM both teased major advances in QC, recently. [35] Rigetti Computing is a startup focused on QC, with applications to AI. They are building a 19 qubit quantum computer to run clustering algorithms. [36] [37] While much R&D remains to create a practical application of their computer and algorithms, investors are still interested. “The company [Rigetti Computing] has raised around $70 million from investors including Andreessen Horowitz, one of Silicon Valley’s most prominent firms.” [38]

As quantum computers become more capable with the addition of more qubits and customized software, we can expect more AI experts to seek out their unique capabilities. Visionary venture capital firms such as BootstrapLabs have held special conference sessions on the opportunities that QC presents to AI. [39]

Neuromorphic

For all that they can accomplish, traditional computers still do not mimic the human mind. CPUs operate in a sequential manner, and, at best, GPUs operate via parallel processing. The mind, however, works highly nonlinearly. Each of our roughly 87 billion neurons has up to 10,000 connections to other nerve cells, externally, and internally hosts hundreds of thousands of coordinated parallel processes (mediated by millions of protein and nucleic acid molecular machines). We routinely use almost all of our neurons in countless waves of electrical signals to process information, all while consuming only tens of watts (compared to thousands of watts for a standard computer) [40] . A chip mimicking the brain’s architecture has the potential to be both more efficient and consume far less power than even the most advanced present-day systems. Neuromorphic computing is one approach for pursuing these advantages.

In a 2014 publication in Science , IBM described a new, ‘brain-like’ approach to IC design [41] . This represents a dramatic jump in scale and flexibility beyond the small processors featuring neural net-based circuitry that have been integrated within computer vision systems and other pattern recognition-heavy applications. [42] It is also fundamentally different from neural nets implemented within FPGA, since neuromorphic chips are optimized for neural net architecture. “The new chip, called TrueNorth, marks a radical departure in chip design and promises to make computers better able to handle complex tasks such as image and voice recognition-jobs at which conventional chips struggle…Each TrueNorth chip contains 5.4 billion transistors wired into an array of 1 million neurons and 256 million synapses. Efforts are already underway to tile more than a dozen such chips together.” [43] IBM also coded entirely new software and encourages computer scientists to create new applications with their neuromorphic chip. In very recent research, investigators at NIST have integrated superconducting Josephson junctions [44] (featured in the D-Wave quantum annealers) into neuromorphic chips delivering a blistering 1GHZ firing rate using as little as 1 atto-Joule in a 3-dimensional architecture. [45]

Another device, the memristor, has been demonstrated to have advantages in power and speed for neuromorphic architectures. These 70 nanometer wide devices were integrated to provide 10 trillion (1E12) operations per watt (compared with a CMOS [46] benchmark approximately three orders of magnitude smaller), engaged in solving the Traveling Salesman Problem. [47]

Going to the next level, Intel announced in September 2017 a neuromorphic chip designed explicitly for AI. [48] “Intel's Loihi chip has 1,024 artificial neurons, or 130,000 simulated neurons with 130 million possible synaptic connections. That's a bit more complex than, say, a lobster's brain, but a long ways from our 80 billion neurons…By simulating this behavior with the Loihi chip, it can (in theory) speed up machine learning while reducing power requirements by up to 1,000 times. What's more, all the learning can be done on-chip, instead of requiring enormous datasets. If incorporated into a computer, such chips could also learn new things on their own, rather than remaining ignorant of tasks they haven't been taught specifically.” [49]

As with other non-von Neumann architectures, software must be developed from scratch to enable the capabilities of neuromorphic computing. Companies such as Applied Brain Research are developing software such as Nengo and Spaun to simulate the mind, while the military and private companies such as Inilabs , sponsors of the Misha Mahowald Prize for Neuromorpic Engineering, and PegusApps are targeting this software gap as an opportunity in machine vision and acoustic pattern recognition . [50] [51] [52] [53]

It remains to be seen how advantageous mimicking the mind is for AI. Neuromorphic computing shows promise, but major advances are still needed in hardware (to provide more ‘neurons’ and their connections) and compatible software.

Conclusions from Part I

There is now no longer a single Moore’s Law, but rather a suite of them, with each new law a function of compute architecture, software and application. Major impetuses for this semiconductor speciation are the intense power and speed demands from AI. As more hardware and software tools become available to AI researchers, specialization will inevitably occur. GPUs, ASICs and FPGAs lead the pack now in terms of AI commercial applications, but if sufficient advances are made in quantum computers and neuromorphic computers we may see wholly new, currently inconceivable applications of AI. [54] This promise makes it potentially timely to invest early in some of these technologies. [55]

Part II of this series will explore further emerging compute technologies such as nanomaterials, optical computing, and DNA and other biological compute strategies. Such capabilities may influence future generations of commercial AI.

Notes

[1] Gordon E. Moore, "Cramming more components onto integrated circuits" Electronics, pp. 114-117, April 19, 1965.

[2] The 50th anniversary of Moore’s Law was celebrated in 2016.

[3] Central Processing Unit

[4] A “von Neumann design” refers to a traditional, linear compute CPU.

[5] Graphics Processing Unit

[6] Application Specific Integrated Circuit

[7] Field Programmable Gate Array

[8] As of January 2018. The semiconductor field moves so fast that much of what is written here will be outdated in terms of specific chip and other descriptions within weeks of this publication; however, broad concepts should remain relevant. It should also be noted that we don’t attempt to be comprehensive in terms of companies or specific hardware in this memo.

[9] http://www.nvidia.com/object/what-is-gpu-computing.html , accessed January 2018.

[10] https://www.analyticsvidhya.com/blog/2017/05/gpus-necessary-for-deep-learning/ , accessed January 2018.

[11] https://www.cnet.com/news/nvidias-3000-titan-v-gpu-offers-massive-power-for-machine-learning/ , accessed January 2018.

[12] https://www.fool.com/investing/2017/12/06/nvidia-is-running-away-with-the-gpu-market.aspx , accessed January 2018.

[13] Original equipment manufacturers

[14] https://en.wikipedia.org/wiki/Application-specific_integrated_circuit , accessed January 2018.

[15] Technically, a GPU itself is an ASIC designed for gaming calculations, but we won’t split hairs.

[16] https://www.forbes.com/sites/moorinsights/2017/08/04/will-asic-chips-become-the-next-big-thing-in-ai/#61aabc7211d9 , accessed January 2018.

[17] https://www.nextplatform.com/2017/05/17/first-depth-look-googles-new-second-generation-tpu/ , accessed January 2018.

[18] https://www.theverge.com/circuitbreaker/2017/10/17/16488414/intel-ai-chips-nervana-neural-network-processor-nnp , accessed January 2018.

[19] https://newsroom.intel.com/news-releases/intel-ai-day-news-release/ , accessed January 2018.

[20] https://www.theverge.com/2017/12/8/16750560/tesla-custom-ai-chips-hardware , accessed January 2018.

[21] https://en.wikipedia.org/wiki/Field-programmable_gate_array , accessed January 2018.

[22] https://software.intel.com/en-us/articles/machine-learning-on-intel-fpgas , accessed January 2018.

[23] http://www.newelectronics.co.uk/electronics-technology/fpgas-key-to-machine-learning-and-ai-apps/156849/ , accessed January 2018.

[24] https://seekingalpha.com/article/4101617-intel-fpgas-powers-microsofts-real-time-artificial-intelligence-cloud-platform , accessed January 2018.

[25] https://www.nextplatform.com/2017/03/21/can-fpgas-beat-gpus-accelerating-next-generation-deep-learning/ , accessed January 2018.

[26] https://dl.acm.org/citation.cfm?id=3021740 , accessed January 2018.

[27] https://www.forbes.com/sites/moorinsights/2017/08/28/microsoft-fpga-wins-versus-google-tpus-for-ai/#daa964939045 , accessed January 2018.

[28] https://www.forbes.com/sites/bernardmarr/2017/09/05/how-quantum-computers-will-revolutionize-artificial-intelligence-machine-learning-and-big-data/#2d0bf8df5609 , accessed January 2018.

[29] https://www.quantamagazine.org/the-era-of-quantum-computing-is-here-outlook-cloudy-20180124/ , accessed January 2018.

[30] “Quantum annealing is a metaheuristic for finding the global minimum of a given objective function over a given set of candidate solutions (candidate states), by a process using quantum fluctuations.” https://en.wikipedia.org/wiki/Quantum_annealing , accessed January 2018.

[31] https://blogs.scientificamerican.com/guest-blog/is-it-quantum-computing-or-not/ , accessed January 2018.

[32] https://www.dwavesys.com/press-releases/d-wave-2000q-system-be-installed-quantum-artificial-intelligence-lab-run-google-nasa , accessed January 2018.

[33] https://futurism.com/ibm-announced-50-qubit-quantum-computer/ , accessed January 2018.

[34] https://arstechnica.com/gadgets/2017/12/microsofts-q-quantum-programming-language-out-now-in-preview/?utm_source=Daily+Email&utm_campaign=bfc784b865-EMAIL_CAMPAIGN_2017_12_12&utm_medium=email&utm_term=0_03a4a88021-bfc784b865-248705265 , accessed January 2018.

[35] https://www.ft.com/stream/ab49fad4-07f5-3950-9aa0-87cdd74ce0b8 , accessed January 2018.

[36] https://www.technologyreview.com/s/609804/a-startup-uses-quantum-computing-to-boost-machine-learning/?utm_source=twitter.com&utm_medium=social&utm_content=2017-12-18&utm_campaign=Technology+Review , accessed January 2018.

[37] Clustering is a machine-learning technique used to organize data into similar groups.

[38] Ibid.

[39] https://bootstraplabs.com/artificial-intelligence/applied-ai-conference-2016/ , accessed January 2018; one author [TAC] is a special advisor to BootstrapLabs Group, LLC.

[40] That we only use 10% of our brains is a myth – e.g., http://www.wired.com/2014/07/everything-you-need-to-know-about-the-10-brain-myth-explained-in-60-seconds/ , accessed January 2018. Even sleeping, we use most of our neurons to dream, breathe, digest food, beat our hearts, etc.

[41] P.A. Merolla, et al., “A million spiking-neuron integrated circuit with a scalable communication network and interface,” Science , 345 (6197), pp. 668-673, 8 August 2014.

[42] http://ieeexplore.ieee.org/document/880850/?reload=true , accessed January 2018.

[43] http://science.sciencemag.org/content/345/6197/668 , accessed January 2018.

[44] “The Josephson effect is the phenomenon of supercurrent—i.e., a current that flows indefinitely long without any voltage applied—across a device known as a Josephson junction (JJ), which consists of two superconductors coupled by a weak link.” https://en.wikipedia.org/wiki/Josephson_effect , accessed January 2018.

[45] https://phys.org/news/2018-01-nist-superconducting-synapse-piece-artificial.html , accessed January 2018.

[46] Complementary metal-oxide-semiconductor

[47] https://spectrum.ieee.org/nanoclast/semiconductors/devices/memristordriven-analog-compute-engine-would-use-chaos-to-compute-efficiently , accessed January 2018.

[48] https://newsroom.intel.com/editorials/intels-new-self-learning-chip-promises-accelerate-artificial-intelligence/ , accessed January 2018.

[49] https://www.engadget.com/2017/09/26/intel-loihi-neuromorphic-chip-human-brain/ , accessed January 2018.

[50] http://www.wired.co.uk/article/ai-neuromorphic-chips-brains , accessed January 2018.

[51] https://www.globalsecurity.org/military/library/budget/fy2016/army-peds/0601102a_1_pb_2016.pdf , accessed January 2018.

[52] https://inilabs.com/ , accessed January 2018.

[53] https://www.pegusapps.com/en/services , accessed January 2018.

[54] https://www.theverge.com/2018/1/27/16940002/google-clips-ai-camera-on-sale-today-waitlist ; https://www.prnewswire.com/news-releases/crowdflower-announces-third-wave-of-ai-for-everyone-challenge-winners-300587610.html , accessed January 2018.

[55] For example: “NVIDIA (NASDAQ:NVDA) stock has returned 80.1% so far in 2017, through Dec. 15, continuing its rapid rise that began last year, when the stock returned a whopping 227%.” [ https://www.fool.com/investing/2017/12/18/how-nvidia-crushed-it-in-2017-a-year-in-review.aspx , accessed January 2018.]. This growth is driven primarily by the intense demand for GPUs because of AI. We might anticipate similar stock price surges elsewhere as other advanced architectures come online for AI.

Like what you read? CONTACT FutureGrasp for more information, speaking engagements and/or collaboration