AI with AI: Artificial Intelligence with Andy Ilachinski – Details, episodes & analysis

For the final (for now?) episode of AI with AI, Andy and Dave discuss the latest in AI news and research, including a political declaration from the US Department of State on the responsible military use of AI and autonomy. NATO begins work on an AI certification standard. The IEEE introduces a new program that provides free access to AI ethics and governance standards. Reported in February, but performed in December, A joint Dept of Defense team performed 12 flight tests (over 17 hours) in which AI agents piloted Lockheed Martin's X-62A VISTA, an F-16 variant. Andy provides a run-down of a large number of recent ChatGPT-related stories. Wolfram "explains" how ChatGPT works. Paul Scharre publishes Four Battlegrounds: Power in the Age of AI. And to come full circle, we began this podcast 6 years ago with the story of AlphaGo beating the world champion. So we close the podcast with news that a non-professional Go player, Kellin Pelrine, beat a top AI system 14 games to one having discovered a 'not super-difficult method for humans to beat the machines. A heartfelt thanks to you all for listening over the years!

Andy and Dave discuss the latest in AI news and research, including the update of the Department of Defense Directive 3000.09 on Autonomy in Weapon Systems. NIST releases the first version of its AI Risk Management Framework. The National AI Research Resource (NAIRR) Task Force publishes its final report, in which it details its plans for a national research infrastructure, as well as its request for $2.6 billion over 6 years to fund the initiatives. DARPA announces the Autonomous Multi-domain Adaptive Swarms-of-Swarms (AMASS) program, a much larger effort (aiming for thousands of autonomous entities) than its previous OFFSET program. And finally, from the Naval Postgraduate School's Energy Academic Group, Kristen Fletcher and Marina Lesse join to discuss their research and efforts in autonomous systems and maritime law and policy, to include a discussion about the DoDD 3000.09 update and the high-altitude balloon incident.

https://www.cna.org/our-media/podcasts/ai-with-ai

Andy and Dave discuss the latest in AI news and research, starting with DARPA moving into Phase 2 of its No Manning Required Ship (NOMARS) program, having selected Serco Inc for its Defiant ship design. The UK releases a roadmap on automated vehicles, Connected & Automated Mobility 2025, and describes new legislation that will place liability for the actions of self-driving vehicles onto manufacturers, and not the occupants. The DOD's Chief Digital and AI Office is preparing to roll out Tradewinds, an open solutions marketplace geared toward identifying new technologies and capabilities. The US bans NVIDIA and AMD from selling or exporting certain types of GPUs (mostly for high-end servers) to China and Russia. A report in Nature examines the "reproducibility crisis" involving machine learning in scientific articles, identifying eight types of "data leaks" in research that raise cause for concern. Google introduces a new AI image noise reduction tool that greatly advances the state of the art for low lighting and resolution images, using RawNeRF, which makes use of the previous neural radiance fields approach, but on raw image data. Hakwan Lau and Oxford University Press make available for free In Consciousness We Trust: the Cognitive Neuroscience of Subjective Experience. And Sam Bendett joins Andy and Dave to discuss the latest from Russia's Army 2022 Expo and other recent developments around the globe.

https://www.cna.org/our-media/podcasts/ai-with-ai

Andy and Dave discuss the latest in AI and autonomy news and research, including an announcement that the Federal Trade Commission is exploring rules for cracking down on harmful commercial surveillance and lax data security, with the public having an opportunity to share input during a virtual public form on 8 September 2022. The Electronic Privacy Information Center (EPIC), with help from Caroline Kraczon, releases The State of State AI Policy, a catalog of AI-related bills that states and local governments have passed, introduced or failed during the 2021-2022 legislative season. In robotics, Xiaomi introduces CyberOne, a 5-foot 9-inch robot that can identify "85 types of environmental sounds and 45 classifications of human emotions." Meanwhile at a recent Russian arms fair, Army-2022, a developer showed off a robot dog with a rocket-propelled grenade strapped to its back. NIST updates its AI Risk Management Framework to the second draft, making it available for review and comment. DARPA launches the SocialCyber project, a hybrid-AI project aimed at helping to protect the integrity of open-source code. BigScience launches BLOOM (BigScience Large Open-science Open-access Multilingual Language Model), a "bigger than GPT-3" multilanguage (46) model that a group of over 1,000 AI researchers has created, that anyone can download and tinker with it for free. Researchers at MIT develop artificial synapses that shuttle protons, resulting in synapses 10,000 times faster than biological ones. China's Comprehensive National Science Center claims that it has developed "mind-reading AI" capable of measuring loyalty to the Chinese Communist Party. Researchers at the University of Sydney demonstrate that human brains are better at identifying deepfakes than people, by examining results directly from neural activity. Researchers at the University of Glasgow combine AI with human vision to see around corners, reconstructing 16x16-pixel images of simple objects that the observer could not directly see. GoogleAI publishes research on Minerva, using language models to solve quantitative reasoning problems, and dramatically increasing the SotA. Researchers from MIT, Columbia, Harvard, and Waterloo publish work on a neural network that solves, explains, and generates university math problems "at a human level." CSET makes available the Country Activity Tracker for AI, an interactive tool on tech competitiveness and collaboration. And a group of researchers at Merced's Cognitive and Information Sciences Program make available Neural Networks in Cognitive Science.

https://www.cna.org/our-media/podcasts/ai-with-ai

Andy and Dave discuss the latest in AI news and research, including an announcement from DeepMind that it is freely providing a database of 200+ million protein structures as predicted by AlphaFold. Researchers at the Max Planck Institute for Intelligent Systems demonstrate how a robot dog can learn to walk in about one hour using a Bayesian optimization algorithm. A chess-playing robot breaks the finger of a seven-year-old boy during a chess match in Moscow. A bill with the Senate Armed Services Committee would require the Department of Defense to accelerate the fielding of new technology to defeat drone swarms. The Chief of Naval Operations Navigation Plan 2022 aims to add 150 uncrewed vessels by 2045. The text-to-image transformer DALL-E is now available in beta. Researchers at Columbia University use an algorithm to identify possible state variables from the observation of systems (such as a double pendulum) and discover "alternate physics"; the algorithm discovers the intrinsic dimension of the observed dynamics and identifies a candidate set of state variables, but in most cases, the scientists found it difficult (if not impossible) to decode those variables to known phenomena. Wolfram Media and Etienne Bernard make Introduction to Machine Learning: Mathematica available for free. And Jeff Edmonds and Sam Bendett join for a discussion on their latest report, Russian Military Autonomy in Ukraine: Four Months In – a closer look at the use of unmanned systems by both Russia and Ukraine.

https://www.cna.org/our-media/podcasts/ai-with-ai

Dr. Anya Fink from CNA's Russia Studies program joins the podcast to discuss the impacts of global sanctions on Russia's technology and AI sector.

Report

CNA: A Technological Divorce: The impact of sanctions and the end of cooperation on Russia's technology and AI sector.

Andy and Dave discuss the latest in AI news and research, including an update from DARPA on its Machine Common Sense program, demonstrating rapidly adapting to changing terrain, carrying dynamic loads, and understanding how to grasp objects [0:55]. The Israeli military fields new tech from Camero-Tech that allows operators to 'see through walls,' using pulse-based ultra-wideband micro-power radar in combination with an AI-based algorithm for tracking live targets [5:01]. In autonomous shipping [8:13], the Suzaka, a cargo ship powered by Orca AI, makes a nearly 500-mile voyage "without human intervention" for 99% of the trip; the Prism Courage sails from the Gulf of Mexico to South Korea "controlled mostly" by HiNAS 2.0, a system by Avikus, a subsidiary of Hyundai; and Promare's and IBM's Mayflower Autonomous Ship travels from the UK to Nova Scotia. In large language models [10:09], a Chinese research team unveils a 174 trillion parameter model, Bagualu ('alchemist pot') and claims it runs an AI model as sophisticated as a human brain (not quite, though); Meta releases the largest open-source AI language model, with OPT-66B, a 66 billion parameter model; and Russia's Yandex opens its 100 billion parameters YaLM to public access. Researchers from the University of Chicago publish a model that can predict future crimes "one week in advance with about 90% accuracy" (referring to general crime levels, not specific people and exact locations), and also demonstrate the potential effects of bias in police response and enforcement [13:32]. In a similar vein, researchers from Berkeley, MIT, and Oxford publish attempts to forecast future world events using the neural network system Autocast, and show that forecasting performance still comes in far below a human expertise baseline [16:37]. Angelo Cangelosi and Minoru Asada provide the (graduate) book of the week, with Cognitive Robotics. 

Andy and Dave discuss the latest in AI news and research, starting with the Department of Defense releasing its Responsible AI Strategy. In the UK, the Ministry of Defence publishes its Defence AI Strategy. The Federal Trade Commission warns policymakers about relying on AI to combat online problems and instead urges them to develop legal frameworks to ensure AI tools do not cause additional harm. YouTuber Yannic Kilcher trains an AI on 4chan's "infamously toxic" Politically Incorrect board, creating a predictably toxic bot, GPT-4chan; he then uses the bot to generate 15,000 posts on the board, quickly receiving condemnation from the academic community. Google suspends and then fires an engineer who claimed that one of its chatbots, LaMDA, achieving sentience; former Google employees Gebru and Mitchell write an opinion piece saying they warned this would happen. For the Fun Site of the Week, a mini version of DALL-E comes to Hugging Face. And finally, IBM researcher Kush Varshney joins Andy and Dave to discuss his book, Trustworthy Machine Learning, which provides AI researchers with practical tools and concepts when developing machine learning systems.

Visit our CNA.org to explore the links mentioned in this episode. 

CNA colleagues Kaia Haney and Heather Roff join Andy and Dave to discuss Responsible AI. They discuss the recent Inclusive National Security seminar on AI and National Security: Gender, Race, and Algorithms. The keynote speaker, Elizabeth Adams spoke on the challenges that society faces in integrating AI technologies in an inclusive fashion, and she identified ways in which consumers of AI-enabled products can ask questions and engage on the topic of inclusivity and bias. The group also discusses a variety of topics around the many challenges that organizations face in operationalizing these ideas, including a revisit of the findings from recent medical research, which found an algorithm was able to identify the race of a subject from x-rays and CAT scans, even with identifying features removed.

Inclusive National Security Series: AI and National Security: Gender, Race and Algorithms

Inclusive National Security webpage

Sign up for the InclusiveNatSec mailing list here.  

Andy and Dave discuss the latest in AI news and research, starting with an announcement that DoD will be updating its Directive 3000.09 on "Autonomous Weapons," with the new Emerging Capabilities Policy Office leading the way [1:25]. The DoD names Diane Staheli as the new chief for Responsible AI [5:19]. NATO launches an AI strategic initiative, Horizon Scanning, to better understand AI and its potential military implications [6:31]. China unveils an autonomous drone carrier ship even though Dave wonders about the use of the terms unmanned and autonomous [8:59]. Stanford University and the Human-Centered AI Center build on their initiative for foundation models by releasing a call to the community for developing norms on the release of foundation models [10:42]. DECIDE-AI continues to develop its reporting guidelines for early-stage clinical evaluation of AI decision support systems [14:39]. The Army successfully demonstrates four waves of seven drones, launched by a single operator, during EDGE 22 [18:31]. Researchers from Zhejiang University and Hong Kong University of S&T demonstrate a swarm of physical micro flying robots, fully autonomous, able to navigate and communicate as a swarm, with fully onboard perception, localization, and control [19:58]. Google Research introduces a new text-to-image generator, Imagen, which uses diffusion models to increase the size and photorealism of an image [24:20]. Researchers discover that an AI algorithm can identify race from X-ray and CT images, even when correcting for variations such as body-mass index but can't explain why or how [31:21]. And Sonantic uses AI to create the voice lines for Val Kilmer in the new movie Top Gun: Maverick [34:18].

RSVP for AI and National Security: Gender, Race, and Algorithms at 12:00 pm on June 7.