New Research: LMArena is rigged Machine Learning Street Talk 2026-09-20 | New Research: LMArena is rigged
Its Not About Scale, Its About Abstraction Machine Learning Street Talk 2024-10-12 | François Chollet discusses the limitations of Large Language Models (LLMs) and proposes a new approach to advancing artificial intelligence. He argues that current AI systems excel at pattern recognition but struggle with logical reasoning and true generalization. This was Chollet's keynote talk at AGI-24, filmed in high-quality. We will be releasing a full interview with him shortly. A teaser clip from that is played in the intro!Chollet introduces the Abstraction and Reasoning Corpus (ARC) as a benchmark for measuring AI progress towards human-like intelligence. He explains the concept of abstraction in AI systems and proposes combining deep learning with program synthesis to overcome current limitations. Chollet suggests that breakthroughs in AI might come from outside major tech labs and encourages researchers to explore new ideas in the pursuit of artificial general intelligence.MLST is sponsored by Tufa Labs:Are you interested in working on ARC and cutting-edge AI research with the MindsAI team (current ARC winners)?Focus: ARC, LLMs, test-time-compute, active inference, system2 reasoning, and more.Future plans: Expanding to complex environments like Warcraft 2 and Starcraft 2.Interested? Apply for an ML research position: benjamin@tufa.aiTOC1. LLM Limitations and Intelligence Concepts[00:00:00] 1.1 LLM Limitations and Composition[00:12:05] 1.2 Intelligence as Process vs. Skill[00:17:15] 1.3 Generalization as Key to AI Progress2. ARC-AGI Benchmark and LLM Performance[00:19:59] 2.1 Introduction to ARC-AGI Benchmark[00:20:05] 2.2 Introduction to ARC-AGI and the ARC Prize[00:23:35] 2.3 Performance of LLMs and Humans on ARC-AGI3. Abstraction in AI Systems[00:26:10] 3.1 The Kaleidoscope Hypothesis and Abstraction Spectrum[00:30:05] 3.2 LLM Capabilities and Limitations in Abstraction[00:32:10] 3.3 Value-Centric vs Program-Centric Abstraction[00:33:25] 3.4 Types of Abstraction in AI Systems4. Advancing AI: Combining Deep Learning and Program Synthesis[00:34:05] 4.1 Limitations of Transformers and Need for Program Synthesis[00:36:45] 4.2 Combining Deep Learning and Program Synthesis[00:39:59] 4.3 Applying Combined Approaches to ARC Tasks[00:44:20] 4.4 State-of-the-Art Solutions for ARCShownotes (new!): dropbox.com/scl/fi/i7nsyoahuei6np95lbjxw/CholletKeynote.pdf?rlkey=t3502kbov5exsdxhderq70b9i&st=1ca91ewz&dl=0[0:01:15] Abstraction and Reasoning Corpus (ARC): AI benchmark (François Chollet)arxiv.org/abs/1911.01547[0:05:30] Monty Hall problem: Probability puzzle (Steve Selvin)tandfonline.com/doi/abs/10.1080/00031305.1975.10479121[0:06:20] LLM training dynamics analysis (Tirumala et al.)arxiv.org/abs/2205.10770[0:10:20] Transformer limitations on compositionality (Dziri et al.)arxiv.org/abs/2305.18654[0:10:25] Reversal Curse in LLMs (Berglund et al.)arxiv.org/abs/2309.12288[0:19:25] Measure of intelligence using algorithmic information theory (François Chollet)arxiv.org/abs/1911.01547[0:20:10] ARC-AGI: GitHub repository (François Chollet)github.com/fchollet/ARC-AGI[0:22:15] ARC Prize: $1,000,000+ competition (François Chollet)arcprize.org[0:33:30] System 1 and System 2 thinking (Daniel Kahneman)amazon.com/Thinking-Fast-Slow-Daniel-Kahneman/dp/0374533555[0:34:00] Core knowledge in infants (Elizabeth Spelke)harvardlds.org/wp-content/uploads/2017/01/SpelkeKinzler07-1.pdf[0:34:30] Embedding interpretive spaces in ML (Tennenholtz et al.)arxiv.org/abs/2310.04475[0:44:20] Hypothesis Search with LLMs for ARC (Wang et al.)arxiv.org/abs/2309.05660[0:44:50] Ryan Greenblatt's high score on ARC public leaderboardarcprize.org
Why does the Chinese Room still haunt AI? Machine Learning Street Talk 2024-10-11 | Dr. Keith Duggar and Dr. Tim Scarfe in this second edition of the hosts-only "philosophical steakhouse" discuss artificial intelligence, consciousness, and understanding. They explore the computational limits of language models, the challenges of training AI systems to be Turing-complete, and the implications of these limitations for AI capabilities.The conversation covers philosophical arguments about machine consciousness, including John Searle's famous Chinese Room thought experiment. They discuss different views on what's required for a system to truly understand or be conscious, touching on ideas from various philosophers and scientists.The hosts also talk about the recent Nobel Prize awarded for work in deep learning, debating its merits and controversies. They touch on the recent Liron Doom Debates show which Duggar was just on, and at the end Ethics vs AI risk. TOC:1. Computational Foundations of AI00:00:00 1.1 Turing Completeness and Computational Limits of Language Models00:06:37 1.2 Finite State Automata vs. Turing Machines in AI00:13:16 1.3 Challenges in Training Turing-Complete Systems00:17:23 1.4 Mapping Turing Machine Programs to Language Models00:20:41 1.5 Future Directions: Hybrid Systems and Novel Programming Approaches2. AI Consciousness and Understanding00:31:30 2.1 Chinese Room Argument and AI Consciousness00:41:15 2.2 Searle's Views on Physical Realization and Consciousness00:50:30 2.3 Emergence and Computational Limitations in Cognitive Science00:59:55 2.4 Friston's Theory on Self-Awareness in Machines01:04:25 2.5 Causal Structures and Understanding in AI Systems01:06:50 2.6 Concept Role Semantics and Language Models01:09:41 2.7 Consciousness and Computational Theories3. AI Impact and Ethics01:21:38 3.1 Deep Learning and the Nobel Prize for Hinton01:29:56 3.2 AI Harm and Existential Risk01:34:05 3.3 Balancing AI Ethics and Practical PolicyRefs:Keith/Liron discussion on Doom Debates (must watch!)youtube.com/watch?v=4v-Qh3JQ4JcLiron's show was in response to our previous "steak house" chat- Is o1 reasoning? youtube.com/watch?v=nO6sDk6vO0g&t=0sMinds brains and programs (Searle)https://home.csulb.edu/~cwallis/382/readings/482/searle.minds.brains.programs.bbs.1980.pdfSearle Google talkyoutube.com/watch?v=rHKwIYsPXLgJ. Mark Bishop on MLSTyoutube.com/watch?v=_KVAzAzO5HUChinese room argumenthttps://plato.stanford.edu/entries/chinese-room/Dancing with pixies - J. Mark Bishopphilarchive.org/rec/BISDWP-2Artificial Intelligence Is Stupid and Causal Reasoning Will Not Fix It - Bishopfrontiersin.org/journals/psychology/articles/10.3389/fpsyg.2020.513474/fullDeconstructing the AI Myth: Fallacies and Harms of Algorithmification (Dagmar Monnett)researchgate.net/publication/382802495_Deconstructing_the_AI_Myth_Fallacies_and_Harms_of_AlgorithmificationNestedly Recursive Functions (Stephen Wolfram)writings.stephenwolfram.com/2024/09/nestedly-recursive-functionsMeasure of intelligence - Cholletarxiv.org/abs/1911.01547What is the philosophy of information - Floridiresearchgate.net/publication/225070243_What_Is_the_Philosophy_of_InformationGodel, Escher, Bach: An Eternal Golden Braidamazon.co.uk/Godel-Escher-Bach-Eternal-Golden/dp/0465026567Shownotes/transcript/refs (autogenerated)dropbox.com/scl/fi/jc6qwn7m2i69t13bo2hlm/TimKeithPSH2.pdf?rlkey=wtegece4e32v4xhvp9expnpcq&st=1tm6868m&dl=0xRecorded Friday 11th Oct 2024
Bold AI Predictions From Cohere Co-founder Machine Learning Street Talk 2024-10-10 | Ivan Zhang, co-founder of Cohere, discusses the company's enterprise-focused AI solutions. He explains Cohere's early emphasis on embedding technology and training models for secure environments. Zhang highlights their implementation of Retrieval-Augmented Generation in healthcare, significantly reducing doctor preparation time. He explores the shift from monolithic AI models to heterogeneous systems and the importance of improving various AI system components. Zhang shares insights on using synthetic data to teach models reasoning, the democratization of software development through AI, and how his gaming skills transfer to running an AI company. He advises young developers to fully embrace AI technologies and offers perspectives on AI reliability, potential risks, and future model architectures.cohere.comivanzhang.cahttps://x.com/1vnzhTOC:00:00:00 Intro00:03:20 AI & Language Model Evolution00:06:09 Future AI Apps & Development00:09:29 Impact on Software Dev Practices00:13:03 Philosophical & Societal Implications00:16:30 Compute Efficiency & RAG00:20:39 Adoption Challenges & Solutions00:22:30 GPU Optimization & Kubernetes Limits00:24:16 Cohere's Implementation Approach00:28:13 Gaming's Professional Influence00:34:45 Transformer Optimizations00:36:45 Future Models & System-Level Focus00:39:20 Inference-Time Computation & Reasoning00:42:05 Capturing Human Thought in AI00:43:15 Research, Hiring & Developer AdviceREFS:00:02:40 The Transformer architecture, arxiv.org/abs/1706.0376200:03:22 The Innovator's Dilemma, amazon.com/Innovators-Dilemma-Technologies-Management-Innovation/dp/163369178000:09:15 The actor model, en.wikipedia.org/wiki/Actor_model00:14:35 John Searle's Chinese Room Argument, https://plato.stanford.edu/entries/chinese-room/00:18:00 Retrieval-Augmented Generation, arxiv.org/abs/2005.1140100:18:40 Retrieval-Augmented Generation, docs.cohere.com/v2/docs/retrieval-augmented-generation-rag00:35:39 Let’s Verify Step by Step, arxiv.org/pdf/2305.2005000:39:20 Adaptive Inference-Time Compute, arxiv.org/abs/2410.0272500:43:20 Ryan Greenblatt ARC entry, redwoodresearch.substack.com/p/getting-50-sota-on-arc-agi-with-gptDisclaimer: This show is part of our Cohere partnership series.
Open-Ended AI: The Key to Superhuman Intelligence? Machine Learning Street Talk 2024-10-04 | Prof. Tim Rocktäschel, AI researcher at UCL and Google DeepMind, talks about open-ended AI systems. These systems aim to keep learning and improving on their own, like evolution does in nature. TOC:00:00:00 Introduction to Open-Ended AI and Key Concepts00:01:37 Tim Rocktäschel's Background and Research Focus00:06:25 Defining Open-Endedness in AI Systems00:10:39 Subjective Nature of Interestingness and Learnability00:16:22 Open-Endedness in Practice: Examples and Limitations00:17:50 Assessing Novelty in Open-ended AI Systems00:20:05 Adversarial Attacks and AI Robustness00:24:05 Rainbow Teaming and LLM Safety00:25:48 Open-ended Research Approaches in AI00:29:05 Balancing Long-term Vision and Exploration in AI Research00:37:25 LLMs in Program Synthesis and Open-Ended Learning00:37:55 Transition from Human-Based to Novel AI Strategies00:39:00 Expanding Context Windows and Prompt Evolution00:40:17 AI Intelligibility and Human-AI Interfaces00:46:04 Self-Improvement and Evolution in AI SystemsShow notes (New!) dropbox.com/scl/fi/5avpsyz8jbn4j1az7kevs/TimR.pdf?rlkey=pqjlcqbtm3undp4udtgfmie8n&st=x50u1d1m&dl=0REFS:00:01:47 - UCL DARK Lab (Rocktäschel) - AI research lab focusing on RL and open-ended learning - ucldark.com00:02:31 - GENIE (Bruce) - Generative interactive environment from unlabelled videos - arxiv.org/abs/2402.1539100:02:42 - Promptbreeder (Fernando) - Self-referential LLM prompt evolution - arxiv.org/abs/2309.1679700:03:05 - Picbreeder (Secretan) - Collaborative online image evolution - dl.acm.org/doi/10.1145/1357054.135732800:03:14 - Why Greatness Cannot Be Planned (Stanley) - Book on open-ended exploration - amazon.com/Why-Greatness-Cannot-Planned-Objective/dp/331915523700:04:36 - NetHack Learning Environment (Küttler) - RL research in procedurally generated game - arxiv.org/abs/2006.1376000:07:35 - Open-ended learning (Clune) - AI systems for continual learning and adaptation - arxiv.org/abs/1905.1098500:07:35 - OMNI (Zhang) - LLMs modeling human interestingness for exploration - arxiv.org/abs/2306.0171100:10:42 - Observer theory (Wolfram) - Computationally bounded observers in complex systems - writings.stephenwolfram.com/2023/12/observer-theory00:15:25 - Human-Timescale Adaptation (Rocktäschel) - RL agent adapting to novel 3D tasks - arxiv.org/abs/2301.0760800:16:15 - Open-Endedness for AGI (Hughes) - Importance of open-ended learning for AGI - arxiv.org/abs/2406.0426800:16:35 - POET algorithm (Wang) - Open-ended approach to generate and solve challenges - arxiv.org/abs/1901.0175300:17:20 - AlphaGo (Silver) - AI mastering the game of Go - https://deepmind.google/technologies/alphago/00:20:35 - Adversarial Go attacks (Dennis) - Exploiting weaknesses in Go AI systems - ifaamas.org/Proceedings/aamas2024/pdfs/p1630.pdf00:22:00 - Levels of AGI (Morris) - Framework for categorizing AGI progress - arxiv.org/abs/2311.0246200:24:30 - Rainbow Teaming (Samvelyan) - LLM-based adversarial prompt generation - arxiv.org/abs/2402.1682200:25:50 - Why Greatness Cannot Be Planned (Stanley) - 'False compass' and 'stepping stone collection' concepts - amazon.com/Why-Greatness-Cannot-Planned-Objective/dp/331915523700:27:45 - AI Debate (Khan) - Improving LLM truthfulness through debate - https://proceedings.mlr.press/v235/khan24a.html00:29:40 - Gemini (Google DeepMind) - Advanced multimodal AI model - https://deepmind.google/technologies/gemini/00:30:15 - How to Take Smart Notes (Ahrens) - Effective note-taking methodology - amazon.com/How-Take-Smart-Notes-Nonfiction/dp/154286650200:35:05 - Voyager (Wang) - Open-ended embodied agent using GPT-4 in Minecraft - arxiv.org/abs/2305.1629100:38:00 - AlphaGo Nature paper (Silver) - Deep neural networks and tree search for Go - nature.com/articles/nature1696100:38:05 - AlphaStar (Vinyals) - AI achieving grandmaster level in StarCraft II - nature.com/articles/s41586-019-1724-z00:42:00 - The Beginning of Infinity (Deutsch) - Book on explanations and scientific progress - amazon.com/Beginning-Infinity-Explanations-Transform-World/dp/014312135900:43:30 - AI model collapse (Shumailov) - Risks of training on AI-generated content - nature.com/articles/s41586-024-07566-y00:48:35 - Chain-of-Thought Prompting (Wei) - Improving LLM reasoning through prompting - arxiv.org/abs/2201.1190300:49:35 - Self-improving neural networks (Schmidhuber) - Early work on self-referential networks - http://people.idsia.ch/~juergen/selfrefnn.html00:54:45 - UCL DARK Lab (UCL Computer Science) - RL and Deep Learning research group - ucl.ac.uk/computer-science/news-and-events/inaugural-lecture-series/inaugural-lecture-professor-tim-rocktaschel
AGI in 5 Years? Ben Goertzel on Superintelligence Machine Learning Street Talk 2024-10-01 | Ben Goertzel discusses AGI development, transhumanism, and the potential societal impacts of superintelligent AI. He predicts human-level AGI by 2029 and argues that the transition to superintelligence could happen within a few years after. Goertzel explores the challenges of AI regulation, the limitations of current language models, and the need for neuro-symbolic approaches in AGI research. He also addresses concerns about resource allocation and cultural perspectives on transhumanism.TOC:[00:00:00] AGI Timeline Predictions and Development Speed[00:00:45] Limitations of Language Models in AGI Development[00:02:18] Current State and Trends in AI Research and Development[00:09:02] Emergent Reasoning Capabilities and Limitations of LLMs[00:18:15] Neuro-Symbolic Approaches and the Future of AI Systems[00:20:00] Evolutionary Algorithms and LLMs in Creative Tasks[00:21:25] Symbolic vs. Sub-Symbolic Approaches in AI[00:28:05] Language as Internal Thought and External Communication[00:30:20] AGI Development and Goal-Directed Behavior[00:35:51] Consciousness and AI: Expanding States of Experience[00:48:50] AI Regulation: Challenges and Approaches[00:55:35] Challenges in AI Regulation[00:59:20] AI Alignment and Ethical Considerations[01:09:15] AGI Development Timeline Predictions[01:12:40] OpenCog Hyperon and AGI Progress[01:17:48] Transhumanism and Resource Allocation Debate[01:20:12] Cultural Perspectives on Transhumanism[01:23:54] AGI and Post-Scarcity Society[01:31:35] Challenges and Implications of AGI DevelopmentNew! PDF Show notes: dropbox.com/scl/fi/fyetzwgoaf70gpovyfc4x/BenGoertzel.pdf?rlkey=pze5dt9vgf01tf2wip32p5hk5&st=svbcofm3&dl=0Refs:00:00:15 Ray Kurzweil's AGI timeline prediction, Ray Kurzweil, en.wikipedia.org/wiki/Technological_singularity00:01:45 Ben Goertzel: SingularityNET founder, Ben Goertzel, singularitynet.io00:02:35 AGI Conference series, AGI Conference Organizers, agi-conf.org/202400:03:55 Ben Goertzel's contributions to AGI, Wikipedia contributors, en.wikipedia.org/wiki/Ben_Goertzel00:11:05 Chain-of-Thought prompting, Subbarao Kambhampati, arxiv.org/abs/2405.0477600:11:35 Algorithmic information content, Pieter Adriaans, https://plato.stanford.edu/entries/information-entropy/00:12:10 Turing completeness in neural networks, Various contributors, https://plato.stanford.edu/entries/turing-machine/00:16:15 AlphaGeometry: AI for geometry problems, Trieu, Li, et al., nature.com/articles/s41586-023-06747-500:18:25 Shane Legg and Ben Goertzel's collaboration, Shane Legg, en.wikipedia.org/wiki/Shane_Legg00:20:00 Evolutionary algorithms in music generation, Yanxu Chen, arxiv.org/html/2409.03715v100:22:00 Peirce's theory of semiotics, Charles Sanders Peirce, https://plato.stanford.edu/entries/peirce-semiotics/00:28:10 Chomsky's view on language, Noam Chomsky, chomsky.info/1983____00:34:05 Greg Egan's 'Diaspora', Greg Egan, amazon.co.uk/Diaspora-post-apocalyptic-thriller-perfect-MIRROR/dp/057508209700:40:35 'The Consciousness Explosion', Ben Goertzel & Gabriel Axel Montes, amazon.com/Consciousness-Explosion-Technological-Experiential-Singularity/dp/B0D8C7QYZD00:41:55 Ray Kurzweil's books on singularity, Ray Kurzweil, amazon.com/Singularity-Near-Humans-Transcend-Biology/dp/014303788900:50:50 California AI regulation bills, California State Senate, sd18.senate.ca.gov/news/senate-unanimously-approves-senator-padillas-artificial-intelligence-package00:56:40 Limitations of Compute Thresholds, Sara Hooker, arxiv.org/abs/2407.0569400:56:55 'Taming Silicon Valley', Gary F. Marcus, penguinrandomhouse.com/books/768076/taming-silicon-valley-by-gary-f-marcus01:09:15 Kurzweil's AGI prediction update, Ray Kurzweil, theguardian.com/technology/article/2024/jun/29/ray-kurzweil-google-ai-the-singularity-is-nearer01:14:45 OpenCog Hyperon framework, Ben Goertzel et al., arxiv.org/abs/2310.1831801:18:25 Malnutrition in Ethiopia, Abriham Shiferaw Areba, frontiersin.org/journals/nutrition/articles/10.3389/fnut.2024.1403591/full01:18:40 Transhumanism ethical debate, Nick Bostrom, nickbostrom.com/papers/history.pdf
This is what happens when you let AIs debate Machine Learning Street Talk 2024-09-27 | Akbir Khan, AI researcher and ICML best paper winner, discusses his work on AI alignment, debate techniques for truthful AI responses, and the future of artificial intelligence.Key points discussed:- Using debate between language models to improve truthfulness in AI responses- Scalable oversight for supervising AI models beyond human-level intelligence- The relationship between intelligence and agency in AI systems- Challenges in AI safety and alignment- The potential for a Cambrian explosion in human-like intelligent systemsThe discussion also explored broader topics:- The wisdom of crowds vs. expert knowledge in machine learning debates- Deceptive alignment and reward tampering in AI systems- Open-ended AI systems and their implications for development and safety- The space of possible minds and defining superintelligence- Cultural evolution and memetics in understanding intelligenceAkbir Khan:https://x.com/akbirkhanhttps://akbir.dev/Show notes and transcript: dropbox.com/scl/fi/sjekivbg3ok6qugsv2p1u/AkbirKhan.pdf?rlkey=ewiyvq0aq7mjvql4u7os0jos2&st=vblhp7af&dl=0TOC (*) are best bits00:00:00 1. Intro: AI alignment and debate techniques for truthful responses *00:05:00 2. Scalable oversight and hidden information settings00:10:05 3. AI agency, intelligence, and progress *00:15:00 4. Base models, RL training, and instrumental goals00:25:11 5. Deceptive alignment and RL challenges in AI *00:30:12 6. Open-ended AI systems and future directions00:35:34 7. Deception, superintelligence, and the space of possible minds *00:40:00 8. Cultural evolution, memetics, and intelligence measurementReferences:1. [00:00:40] Akbir Khan et al. ICML 2024 Best Paper: "Debating with More Persuasive LLMs Leads to More Truthful Answers" arxiv.org/html/2402.06782v32. [00:03:28] Yann LeCun on machine learning debates youtube.com/watch?v=OKkEdTchsiE3. [00:06:05] OpenAI's Superalignment team openai.com/index/introducing-superalignment4. [00:08:10] Sam Bowman on scalable oversight in AI systems arxiv.org/abs/2211.035405. [00:10:35] Sam Bowman on the sandwich protocol alignmentforum.org/posts/nekLYqbCEBDEfbLzF/artificial-sandwiching-when-can-we-test-scalable-alignment6. [00:14:35] Janus' article on "Simulators" and LLMs lesswrong.com/posts/vJFdjigzmcXMhNTsx/simulators7. [00:16:35] Thomas Suddendorf's book "The Gap: The Science of What Separates Us from Other Animals" https://www.amazon.in/GAP-Science-Separates-Other-Animals/dp/04650301498. [00:19:10] DeepMind on responsible AI https://deepmind.google/about/responsibility-safety/9. [00:20:50] Technological singularity en.wikipedia.org/wiki/Technological_singularity10. [00:21:30] Eliezer Yudkowsky on FOOM (Fast takeoff) intelligence.org/files/AIFoomDebate.pdf11. [00:21:45] Sammy Martin on recursive self-improvement in AI alignmentforum.org/posts/5WECpYABCT62TJrhY/will-ai-undergo-discontinuous-progress12. [00:24:25] LessWrong community lesswrong.com13. [00:24:35] Nora Belrose on AI alignment and deception lesswrong.com/posts/YsFZF3K9tuzbfrLxo/counting-arguments-provide-no-evidence-for-ai-doom14. [00:25:35] Evan Hubinger on deceptive alignment in AI systems lesswrong.com/posts/zthDPAjh9w6Ytbeks/deceptive-alignment15. [00:26:50] Anthropic's article on reward tampering in language models anthropic.com/research/reward-tampering16. [00:32:35] Kenneth Stanley's work on open-endedness in AI amazon.co.uk/Why-Greatness-Cannot-Planned-Objective/dp/331915523717. [00:34:58] Ryan Greenblatt, Buck Shlegeris et al. on AI safety protocols arxiv.org/pdf/2312.0694218. [00:37:20] Aaron Sloman's concept of 'the space of possible minds' cs.bham.ac.uk/research/projects/cogaff/sloman-space-of-minds-84.pdf19. [00:38:25] François Chollet on defining and measuring intelligence in AI arxiv.org/abs/1911.0154720. [00:42:30] Richard Dawkins on memetics amazon.co.uk/Selfish-Gene-Richard-Dawkins/dp/019286092521. [00:42:45] Jonathan Cook et al. on Artificial Generational Intelligence arxiv.org/abs/2406.0039222. [00:45:00] Peng on determinants of cryptocurrency pricing emerald.com/insight/content/doi/10.1108/CAFR-05-2023-0053/full/html
Taming Silicon Valley - Prof. Gary Marcus Machine Learning Street Talk 2024-09-24 | AI expert Prof. Gary Marcus doesn't mince words about today's artificial intelligence. He argues that despite the buzz, chatbots like ChatGPT aren't as smart as they seem and could cause real problems if we're not careful.Marcus is worried about tech companies putting profits before people. He thinks AI could make fake news and privacy issues even worse. He's also concerned that a few big tech companies have too much power. Looking ahead, Marcus believes the AI hype will die down as reality sets in. He wants to see AI developed in smarter, more responsible ways. His message to the public? We need to speak up and demand better AI before it's too late.Buy Taming Silicon Valley:amzn.to/3XTlC5sGary Marcus:garymarcus.substack.comhttps://x.com/GaryMarcusInterviewer: Dr. Tim Scarfe(Refs in top comment)TOC[00:00:00] AI Flaws, Improvements & Industry Critique[00:16:29] AI Safety Theater & Image Generation Issues[00:23:49] AI's Lack of World Models & Human-like Understanding[00:31:09] LLMs: Superficial Intelligence vs. True Reasoning[00:34:45] AI in Specialized Domains: Chess, Coding & Limitations[00:42:10] AI-Generated Code: Capabilities & Human-AI Interaction[00:48:10] AI Regulation: Industry Resistance & Oversight Challenges[00:54:55] Copyright Issues in AI & Tech Business Models[00:57:26] AI's Societal Impact: Risks, Misinformation & Ethics[01:23:14] AI X-risk, Alignment & Moral Principles Implementation[01:37:10] Persistent AI Flaws: System Limitations & Architecture Challenges[01:44:33] AI Future: Surveillance Concerns, Economic Challenges & Neuro-Symbolic AI
Rethinking the Mind - Prof. Mark Solms Machine Learning Street Talk 2024-09-18 | Prof. Mark Solms, a neuroscientist and psychoanalyst, discusses his groundbreaking work on consciousness, challenging conventional cortex-centric views and emphasizing the role of brainstem structures in generating consciousness and affect.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. Key points discussed:The limitations of cortex-centric approaches to consciousness studies.Evidence from decorticated animals and hydranencephalic children supporting the brainstem's role in consciousness.The relationship between homeostasis, the free energy principle, and consciousness.Critiques of behaviorism and modern theories of consciousness.The importance of subjective experience in understanding brain function.The discussion also explored broader topics:The potential impact of affect-based theories on AI development.The role of the SEEKING system in exploration and learning.Connections between neuroscience, psychoanalysis, and philosophy of mind.Challenges in studying consciousness and the limitations of current theories.Mark Solms: https://neuroscience.uct.ac.za/contacts/mark-solmsShow notes and transcript: dropbox.com/scl/fo/roipwmnlfmwk2e7kivzms/ACjZF-VIGC2-Suo30KcwVV0?rlkey=53y8v2cajfcgrf17p1h7v3suz&st=z8vu81hn&dl=0TOC (*) are best bits00:00:00 1. Intro: Challenging vision-centric approaches to consciousness *00:02:20 2. Evidence from decorticated animals and hydranencephalic children *00:07:40 3. Emotional responses in hydranencephalic children00:10:40 4. Brainstem stimulation and affective states00:15:00 5. Brainstem's role in generating affective consciousness *00:21:50 6. Dual-aspect monism and the mind-brain relationship00:29:37 7. Information, affect, and the hard problem of consciousness *00:37:25 8. Wheeler's participatory universe and Chalmers' theories00:48:51 9. Homeostasis, free energy principle, and consciousness *00:59:25 10. Affect, voluntary behavior, and decision-making01:05:45 11. Psychoactive substances, REM sleep, and consciousness research01:12:14 12. Critiquing behaviorism and modern consciousness theories *01:24:25 13. The SEEKING system and exploration in neuroscienceRefs:1. Mark Solms' book "The Hidden Spring" [00:20:34] (MUST READ!) amzn.to/3XyETb32. Karl Friston's free energy principle [00:03:50] nature.com/articles/nrn27873. Hydranencephaly condition [00:07:10] en.wikipedia.org/wiki/Hydranencephaly4. Periaqueductal gray (PAG) [00:08:57] en.wikipedia.org/wiki/Periaqueductal_gray5. Positron Emission Tomography (PET) [00:13:52] en.wikipedia.org/wiki/Positron_emission_tomography6. Paul MacLean's triune brain theory [00:03:30] en.wikipedia.org/wiki/Triune_brain7. Baruch Spinoza's philosophy of mind [00:23:48] https://plato.stanford.edu/entries/spinoza-epistemology-mind8. Claude Shannon's "A Mathematical Theory of Communication" [00:32:15] https://people.math.harvard.edu/~ctm/home/text/others/shannon/entropy/entropy.pdf9. Francis Crick's "The Astonishing Hypothesis" [00:39:57] en.wikipedia.org/wiki/The_Astonishing_Hypothesis10. Frank Jackson's Knowledge Argument [00:40:54] https://plato.stanford.edu/entries/qualia-knowledge/11. Mesolimbic dopamine system [01:11:51] en.wikipedia.org/wiki/Mesolimbic_pathway12. Jaak Panksepp's SEEKING system [01:25:23] en.wikipedia.org/wiki/Jaak_Panksepp#Affective_neuroscience
The scientist who coined retrieval augmented generation Machine Learning Street Talk 2024-09-16 | Dr. Patrick Lewis, who coined the term RAG (Retrieval Augmented Generation) and now works at Cohere, discusses the evolution of language models, RAG systems, and challenges in AI evaluation. MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmented generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. Key topics covered:- Origins and evolution of Retrieval Augmented Generation (RAG)- Challenges in evaluating RAG systems and language models- Human-AI collaboration in research and knowledge work- Word embeddings and the progression to modern language models- Dense vs sparse retrieval methods in information retrievalThe discussion also explored broader implications and applications:- Balancing faithfulness and fluency in RAG systems- User interface design for AI-augmented research tools- The journey from chemistry to AI research- Challenges in enterprise search compared to web search- The importance of data quality in training AI modelsPatrick Lewis: patricklewis.ioCohere Command Models, check them out - they are amazing for RAG!cohere.com/commandTOC00:00:00 1. Intro to RAG00:05:30 2. RAG Evaluation: Poll framework & model performance00:12:55 3. Data Quality: Cleanliness vs scale in AI training00:15:13 4. Human-AI Collaboration: Research agents & UI design00:22:57 5. RAG Origins: Open-domain QA to generative models00:30:18 6. RAG Challenges: Info retrieval, tool use, faithfulness00:42:01 7. Dense vs Sparse Retrieval: Techniques & trade-offs00:47:02 8. RAG Applications: Grounding, attribution, hallucination prevention00:54:04 9. UI for RAG: Human-computer interaction & model optimization00:59:01 10. Word Embeddings: Word2Vec, GloVe, and semantic spaces01:06:43 11. Language Model Evolution: BERT, GPT, and beyond01:11:38 12. AI & Human Cognition: Sequential processing & chain-of-thoughtRefs:1. Retrieval Augmented Generation (RAG) paper / Patrick Lewis et al. [00:27:45] arxiv.org/abs/2005.114012. LAMA (LAnguage Model Analysis) probe / Petroni et al. [00:26:35] arxiv.org/abs/1909.010663. KILT (Knowledge Intensive Language Tasks) benchmark / Petroni et al. [00:27:05] arxiv.org/abs/2009.022524. Word2Vec algorithm / Tomas Mikolov et al. [01:00:25] arxiv.org/abs/1301.37815. GloVe (Global Vectors for Word Representation) / Pennington et al. [01:04:35] https://nlp.stanford.edu/projects/glove/6. BERT (Bidirectional Encoder Representations from Transformers) / Devlin et al. [01:08:00] arxiv.org/abs/1810.048057. 'The Language Game' book / Nick Chater and Morten H. Christiansen [01:11:40] amzn.to/4grEUpGDisclaimer: This is the sixth video from our Cohere partnership. We were not told what to say in the interview. Filmed in London in June 2024.
Is o1-preview reasoning? Machine Learning Street Talk 2024-09-15 | Dr. Tim Scarfe and Dr. Keith Duggar discuss OpenAI's new models and their capabilities. They critically analyse claims about AI reasoning, explore the limitations of current language models, and debate the nature of intelligence and computation. Throughout, they emphasize the importance of human oversight in AI applications and discuss potential future developments in the field.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. We edited about an hour off this conversation, see full one on patreon - patreon.com/posts/tim-and-keith-on-112091132TOC:00:00:00 1. Introduction and AI hype cycles00:02:09 2. Computational limits of AI systems00:03:57 3. Neural Networks vs. Turing Machines00:11:55 4. Computational models in AI00:13:03 5. What is Reasoning?00:21:08 6. Chain-of-thought prompting00:26:02 7. AI code generation and complexity00:34:24 8. AI assistance vs. human problem-solving00:35:04 9. Limitations of AI in reasoning and problem-solving00:46:27 10. Knowledge acquisition and inference in AI00:53:36 11. Comparing AI and human reasoning capabilities00:58:58 12. LLMs as cognitive tools01:00:32 13. Testing o1-preview on a logic puzzle01:20:48 14. AI-assisted coding: strengths and limitations
AIs can now imagine video games in real-time Machine Learning Street Talk 2024-09-13 | Ashley Edwards, who was working at DeepMind when she co-authored the Genie paper and is now at Runway, covered several key aspects of the Genie AI system and its applications in video generation, robotics, and game creation. MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. Genie's approach to learning interactive environments, balancing compression and fidelity.The use of latent action models and VQE models for video processing and tokenization.Challenges in maintaining action consistency across frames and integrating text-to-image models.Evaluation metrics for AI-generated content, such as FID and PS&R diff metrics.The discussion also explored broader implications and applications:The potential impact of AI video generation on content creation jobs.Applications of Genie in game generation and robotics.The use of foundation models in robotics and the differences between internet video data and specialized robotics data.Challenges in mapping AI-generated actions to real-world robotic actions.Ashley Edwards: ashedwards.github.ioTOC (*) are best bits00:00:00 1. Intro to Genie & Brave Search API: Trade-offs & limitations *00:02:26 2. Genie's Architecture: Latent action, VQE, video processing *00:05:06 3. Genie's Constraints: Frame consistency & image model integration00:07:26 4. Evaluation: FID, PS&R diff metrics & latent induction methods00:09:44 5. AI Video Gen: Content creation impact, depth & parallax effects00:11:39 6. Model Scaling: Training data impact & computational trade-offs00:13:50 7. Game & Robotics Apps: Gamification & action mapping challenges *00:16:16 8. Robotics Foundation Models: Action space & data considerations *00:19:18 9. Mask-GPT & Video Frames: Real-time optimization, RL from videos00:20:34 10. Research Challenges: AI value, efficiency vs. quality, safety00:24:20 11. Future Dev: Efficiency improvements & fine-tuning strategiesRefs:1. Genie (learning interactive environments from videos) / Ashley and DM collegues [00:01] arxiv.org/abs/2402.153912. VQ-VAE (Vector Quantized Variational Autoencoder) / Aaron van den Oord, Oriol Vinyals, Koray Kavukcuoglu [02:43] arxiv.org/abs/1711.009373. FID (Fréchet Inception Distance) metric / Martin Heusel et al. [07:37] arxiv.org/abs/1706.085004. PS&R (Precision and Recall) metric / Mehdi S. M. Sajjadi et al. [08:02] arxiv.org/abs/1806.000355. Vision Transformer (ViT) architecture / Alexey Dosovitskiy et al. [12:14] arxiv.org/abs/2010.119296. Genie (robotics foundation models) / Google DeepMind [17:34] https://deepmind.google/research/publications/60474/7. Chelsea Finn's lab work on robotics datasets / Chelsea Finn [17:38] https://ai.stanford.edu/~cbfinn/8. Imitation from observation in reinforcement learning / YuXuan Liu [20:58] arxiv.org/abs/1707.033749. Waymo's autonomous driving technology / Waymo [22:38] waymo.com10. Gen3 model release by Runway / Runway [23:48] runwayml.com11. Classifier-free guidance technique / Jonathan Ho and Tim Salimans [24:43] arxiv.org/abs/2207.12598
LLMs in enterprise environments Machine Learning Street Talk 2024-09-12 | Saurabh Baji discusses Cohere's approach to developing and deploying large language models (LLMs) for enterprise use. * Cohere focuses on pragmatic, efficient models tailored for business applications rather than pursuing the largest possible models.* They offer flexible deployment options, from cloud services to on-premises installations, to meet diverse enterprise needs.* Retrieval-augmented generation (RAG) is highlighted as a critical capability, allowing models to leverage enterprise data securely.* Cohere emphasizes model customization, fine-tuning, and tools like reranking to optimize performance for specific use cases.* The company has seen significant growth, transitioning from developer-focused to enterprise-oriented services.* Major customers like Oracle, Fujitsu, and TD Bank are using Cohere's models across various applications, from HR to finance.* Baji predicts a surge in enterprise AI adoption over the next 12-18 months as more companies move from experimentation to production.* He emphasizes the importance of trust, security, and verifiability in enterprise AI applications.The interview provides insights into Cohere's strategy, technology, and vision for the future of enterprise AI adoption.linkedin.com/in/saurabhbajihttps://x.com/sbajicohere.comcohere.com/businessMLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. TOC (*) are best bits00:00:00 1. Introduction and Background00:04:24 2. Cloud Infrastructure and LLM Optimization00:06:43 2.1 Model deployment and fine-tuning strategies *00:09:37 3. Enterprise AI Deployment Strategies00:11:10 3.1 Retrieval-augmented generation in enterprise environments *00:13:40 3.2 Standardization vs. customization in cloud services *00:18:20 4. AI Model Evaluation and Deployment00:18:20 4.1 Comprehensive evaluation frameworks *00:21:20 4.2 Key components of AI model stacks *00:25:50 5. Retrieval Augmented Generation (RAG) in Enterprise00:32:10 5.1 Pragmatic approach to RAG implementation *00:33:45 6. AI Agents and Tool Integration00:33:45 6.1 Leveraging tools for AI insights *00:35:30 6.2 Agent-based AI systems and diagnostics *00:42:55 7. AI Transparency and Reasoning Capabilities00:49:10 8. AI Model Training and Customization00:57:10 9. Enterprise AI Model Management01:02:10 9.1 Managing AI model versions for enterprise customers *01:04:30 9.2 Future of language model programming *01:06:10 10. AI-Driven Software Development01:06:10 10.1 AI bridging human expression and task achievement *01:08:00 10.2 AI-driven virtual app fabrics in enterprise *01:13:33 11. Future of AI and Enterprise Applications01:21:55 12. Cohere's Customers and Use Cases01:21:55 12.1 Cohere's growth and enterprise partnerships *01:27:14 12.2 Diverse customers using generative AI *01:27:50 12.3 Industry adaptation to generative AI *01:29:00 13. Technical Advantages of Cohere Models01:29:00 13.1 Handling large context windows *01:29:40 13.2 Low latency impact on developer productivity *Disclaimer: This is the fifth video from our Cohere partnership. We were not told what to say in the interview, and didn't edit anything out from the interview. Filmed in Seattle in Aug 2024.
David Hansons Vision for Sentient Robots Machine Learning Street Talk 2024-09-10 | David Hanson, CEO of Hanson Robotics and creator of the humanoid robot Sofia, explores the intersection of artificial intelligence, ethics, and human potential. In this thought-provoking interview, Hanson discusses his vision for developing AI systems that embody the best aspects of humanity while pushing beyond our current limitations, aiming to achieve what he calls "super wisdom."MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. The interview with David Hanson covers:The importance of incorporating biological drives and compassion into AI systemsHanson's concept of "existential pattern ethics" as a basis for AI moralityThe potential for AI to enhance human intelligence and wisdomChallenges in developing artificial general intelligence (AGI)The need to democratize AI technologies globallyPotential future advancements in human-AI integration and their societal impactsConcerns about technological augmentation exacerbating inequalityThe role of ethics in guiding AI development and deploymentHanson advocates for creating AI systems that embody the best aspects of humanity while surpassing current human limitations, aiming for "super wisdom" rather than just artificial super intelligence.David Hanson:hansonrobotics.com/david-hansonyoutube.com/watch?v=9u1O954cMmETOC1. Introduction and Background [00:00:00] 1.1. David Hanson's interdisciplinary background [0:01:49] 1.2. Introduction to Sofia, the realistic robot [0:03:27]2. Human Cognition and AI [0:03:50] 2.1. Importance of social interaction in cognition [0:03:50] 2.2. Compassion as distinguishing factor [0:05:55] 2.3. AI augmenting human intelligence [0:09:54]3. Developing Human-like AI [0:13:17] 3.1. Incorporating biological drives in AI [0:13:17] 3.2. Creating AI with agency [0:20:34] 3.3. Implementing flexible desires in AI [0:23:23]4. Ethics and Morality in AI [0:27:53] 4.1. Enhancing humanity through AI [0:27:53] 4.2. Existential pattern ethics [0:30:14] 4.3. Expanding morality beyond restrictions [0:35:35]5. Societal Impact of AI [0:38:07] 5.1. AI adoption and integration [0:38:07] 5.2. Democratizing AI technologies [0:38:32] 5.3. Human-AI integration and identity [0:43:37]6. Future Considerations [0:50:03] 6.1. Technological augmentation and inequality [0:50:03] 6.2. Emerging technologies for mental health [0:50:32] 6.3. Corporate ethics in AI development [0:52:26]This was filmed at AGI-24
The Fabric of Knowledge - David Spivak Machine Learning Street Talk 2024-09-05 | David Spivak, a mathematician known for his work in category theory, discusses a wide range of topics related to intelligence, creativity, and the nature of knowledge. He explains category theory in simple terms and explores how it relates to understanding complex systems and relationships.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. We discuss abstract concepts like collective intelligence, the importance of embodiment in understanding the world, and how we acquire and process knowledge. Spivak shares his thoughts on creativity, discussing where it comes from and how it might be modeled mathematically.A significant portion of the discussion focuses on the impact of artificial intelligence on human thinking and its potential role in the evolution of intelligence. Spivak also touches on the importance of language, particularly written language, in transmitting knowledge and shaping our understanding of the world.Pod: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/The-Fabric-of-Knowledge---David-Spivak-e2o220hDavid Spivak:http://www.dspivak.netTOC:00:00:00 Introduction to category theory and functors00:04:40 Collective intelligence and sense-making00:09:54 Embodiment and physical concepts in knowledge acquisition00:16:23 Creativity, open-endedness, and AI's impact on thinking00:25:46 Modeling creativity and the evolution of intelligence00:36:04 Evolution, optimization, and the significance of AI00:44:14 Written language and its impact on knowledge transmissionREFS:Mike Levin's workscholar.google.com/citations?user=luouyakAAAAJ&hl=enEric Smith's videos on complexity and early lifeyoutube.com/watch?v=SpJZw-68QyERichard Dawkins' book "The Selfish Gene"amzn.to/3X73X8wCarl Sagan's statement about the cosmos knowing itselfamzn.to/3XhPruKHerbert Simon's concept of "satisficing"https://plato.stanford.edu/entries/bounded-rationality/DeepMind paper on open-ended systemsarxiv.org/abs/2406.04268Karl Friston's work on active inferencehttps://direct.mit.edu/books/oa-monograph/5299/Active-InferenceThe-Free-Energy-Principle-in-MindMIT category theory lectures by David Spivak (available on the Topos Institute channel)youtube.com/watch?v=UusLtx9fIjs
Neural and Non-Neural AI, Reasoning, Transformers, and LSTMs Machine Learning Street Talk 2024-08-28 | Jürgen Schmidhuber, the father of generative AI shares his groundbreaking work in deep learning and artificial intelligence. In this exclusive interview, he discusses the history of AI, some of his contributions to the field, and his vision for the future of intelligent machines. Schmidhuber offers unique insights into the exponential growth of technology and the potential impact of AI on humanity and the universe.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. TOC00:00:00 Intro00:03:38 Reasoning00:13:09 Potential AI Breakthroughs Reducing Computation Needs00:20:39 Memorization vs. Generalization in AI00:25:19 Approach to the ARC Challenge00:29:10 Perceptions of Chat GPT and AGI00:58:45 Abstract Principles of Jurgen's Approach01:04:17 Analogical Reasoning and Compression01:05:48 Breakthroughs in 1991: the P, the G, and the T in ChatGPT and Generative AI01:15:50 Use of LSTM in Language Models by Tech Giants01:21:08 Neural Network Aspect Ratio Theory01:26:53 Reinforcement Learning Without Explicit TeachersRefs:★ "Annotated History of Modern AI and Deep Learning" (2022 survey by Schmidhuber):★ Chain Rule For Backward Credit Assignment (Leibniz, 1676)★ First Neural Net / Linear Regression / Shallow Learning (Gauss & Legendre, circa 1800)★ First 20th Century Pioneer of Practical AI (Quevedo, 1914)★ First Recurrent NN (RNN) Architecture (Lenz, Ising, 1920-1925)★ AI Theory: Fundamental Limitations of Computation and Computation-Based AI (Gödel, 1931-34)★ Unpublished ideas about evolving RNNs (Turing, 1948)★ Multilayer Feedforward NN Without Deep Learning (Rosenblatt, 1958)★ First Published Learning RNNs (Amari and others, ~1972)★ First Deep Learning (Ivakhnenko & Lapa, 1965)★ Deep Learning by Stochastic Gradient Descent (Amari, 1967-68)★ ReLUs (Fukushima, 1969)★ Backpropagation (Linnainmaa, 1970); precursor (Kelley, 1960)★ Backpropagation for NNs (Werbos, 1982)★ First Deep Convolutional NN (Fukushima, 1979); later combined with Backprop (Waibel 1987, Zhang 1988).★ Metalearning or Learning to Learn (Schmidhuber, 1987)★ Generative Adversarial Networks / Artificial Curiosity / NN Online Planners (Schmidhuber, Feb 1990; see the G in Generative AI and ChatGPT)★ NNs Learn to Generate Subgoals and Work on Command (Schmidhuber, April 1990)★ NNs Learn to Program NNs: Unnormalized Linear Transformer (Schmidhuber, March 1991; see the T in ChatGPT)★ Deep Learning by Self-Supervised Pre-Training. Distilling NNs (Schmidhuber, April 1991; see the P in ChatGPT)★ Experiments with Pre-Training; Analysis of Vanishing/Exploding Gradients, Roots of Long Short-Term Memory / Highway Nets / ResNets (Hochreiter, June 1991, further developed 1999-2015 with other students of Schmidhuber)★ LSTM journal paper (1997, most cited AI paper of the 20th century)★ xLSTM (Hochreiter, 2024)★ Reinforcement Learning Prompt Engineer for Abstract Reasoning and Planning (Schmidhuber 2015)★ Mindstorms in Natural Language-Based Societies of Mind (2023 paper by Schmidhuber's team) arxiv.org/abs/2305.17066★ Bremermann's physical limit of computation (1982)EXTERNAL LINKSCogX 2018 - Professor Juergen Schmidhuberyoutube.com/watch?v=17shdT9-wuADiscovering Neural Nets with Low Kolmogorov Complexity and High Generalization Capability (Neural Networks, 1997)https://sferics.idsia.ch/pub/juergen/loconet.pdfThe paradox at the heart of mathematics: Gödel's Incompleteness Theorem - Marcus du Sautoyyoutube.com/watch?v=I4pQbo5MQOsThe Philosophy of Science - Hilary Putnam & Bryan Magee (1977)youtube.com/watch?v=JJB2q8ufAgkOptimal Ordered Problem Solverarxiv.org/abs/cs/0207097Levin's Universal Search from 1973rjlipton.com/2011/03/14/levins-great-discoverieshttps://people.idsia.ch/~juergen/optimalsearch.htmlOn Learning to Thinkarxiv.org/abs/1511.09249Mindstorms in Natural Language-Based Societies of Mind Untersuchungen zu dynamischen neuronalen Netzenhttps://www.bioinf.jku.at/publications/older/3804.pdfEvolutionary Principles in Self-Referential Learninghttps://people.idsia.ch/~juergen/diploma1987ocr.pdfHans-Joachim Bremermannen.wikipedia.org/wiki/Bremermann%27s_limitHighway Networksarxiv.org/abs/1505.00387https://people.idsia.ch/~juergen/highway-networks.html The principles of Deep Learning Theoryamzn.to/3WJtPajUnderstanding Deep Learning amzn.to/4doDk63Discovering Problem Solutions with Low Kolmogorov Complexity and High Generalization Capability (ICML 1995)https://sferics.idsia.ch/pub/juergen/icmlkolmogorov.pdf"History of Modern AI and Deep Learning":https://people.idsia.ch/~juergen/deep-learning-history.html
AI should NOT be regulated at all! - Domingos Machine Learning Street Talk 2024-08-25 | Professor Pedro Domingos, is an AI researcher and professor of computer science. He expresses skepticism about current AI regulation efforts and argues for faster AI development rather than slowing it down. He also discusses the need for new innovations to fulfil the promises of current AI techniques.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmented generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. Show notes:* Domingos' views on AI regulation and why he believes it's misguided* His thoughts on the current state of AI technology and its limitations* Discussion of his novel "2040", a satirical take on AI and tech culture* Explanation of his work on "tensor logic", which aims to unify neural networks and symbolic AI* Critiques of other approaches in AI, including those of OpenAI and Gary Marcus* Thoughts on the AI "bubble" and potential future developments in the fieldProf. Pedro Domingos:https://x.com/pmddomingos2040: A Silicon Valley Satire [Pedro's new book]amzn.to/3T51ISdPod version: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/AI-should-NOT-be-regulated-at-all----Prof--Pedro-Domingos-e2njadaTOC:00:00:00 Intro00:06:31 Bio00:08:40 Filmmaking skit00:10:35 AI and the wisdom of crowds00:19:49 Social Media00:27:48 Master algorithm00:30:48 Neurosymbolic AI / abstraction00:39:01 Language00:45:38 Chomsky01:00:49 2040 Book01:18:03 Satire as a shield for criticism?01:29:12 AI Regulation / Gary Marcus01:52:37 Copyright01:56:11 Stochastic parrots come home to roost02:00:03 Privacy02:01:55 LLM ecosystem02:05:06 Tensor logicRefs:The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World [Pedro Domingos]amzn.to/3MiWs9BRebooting AI: Building Artificial Intelligence We Can Trust [Gary Marcus]amzn.to/3AAywvLFlash Boys [Michael Lewis]amzn.to/4dUGm1M
ML Robustness & Engineering - Andrew Ilyas (MIT) Machine Learning Street Talk 2024-08-22 | Andrew Ilyas, a PhD student at MIT who is about to start as a professor at CMU. We discuss Data modeling and understanding how datasets influence model predictions, Adversarial examples in machine learning and why they occur, Robustness in machine learning models, Black box attacks on machine learning systems, Biases in data collection and dataset creation, particularly in ImageNet and Self-selection bias in data and methods to address it.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api This is video 4/13 from ICML 2024Andrew's site:andrewilyas.comhttps://x.com/andrew_ilyasTOC:00:00:00 - Introduction and Andrew's background00:03:52 - Overview of the machine learning pipeline00:06:31 - Data modeling paper discussion00:26:28 - TRAK: Evolution of data modeling work00:43:58 - Discussion on abstraction, reasoning, and neural networks00:53:16 - "Adversarial Examples Are Not Bugs, They Are Features" paper01:03:24 - Types of features learned by neural networks01:10:51 - Black box attacks paper01:15:39 - Work on data collection and bias01:25:48 - Future research plans and closing thoughtsPod version: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Adversarial-Examples-and-Data-Modelling---Andrew-Ilyas-MIT-e2nfov9References:Adversarial Examples Are Not Bugs, They Are Featuresarxiv.org/pdf/1905.02175TRAK: Attributing Model Behavior at Scalearxiv.org/pdf/2303.14186Datamodels: Predicting Predictions from Training Dataarxiv.org/pdf/2202.00622Adversarial Examples Are Not Bugs, They Are Featuresarxiv.org/pdf/1905.02175IMAGENET-TRAINED CNNSarxiv.org/pdf/1811.12231ZOO: Zeroth Order Optimization Based Black-boxarxiv.org/pdf/1708.03999A Spline Theory of Deep Networkshttps://proceedings.mlr.press/v80/balestriero18b/balestriero18b.pdfScaling Monosemanticityhttps://transformer-circuits.pub/2024/scaling-monosemanticity/Adversarial Examples Are Not Bugs, They Are Featuresgradientscience.org/advAdversarial Robustness Limits via Scaling-Law and Human-Alignment Studieshttps://proceedings.mlr.press/v235/bartoldson24a.htmlPrior Convictions: Black-Box Adversarial Attacks with Bandits and Priorsarxiv.org/abs/1807.07978Estimation of Standard Auction Modelsarxiv.org/abs/2205.02060From ImageNet to Image Classification: Contextualizing Progress on Benchmarksarxiv.org/abs/2005.11295Estimation of Standard Auction Modelsarxiv.org/abs/2205.02060What Makes A Good Fisherman? Linear Regression under Self-Selection Biasarxiv.org/abs/2205.03246Towards Tracing Factual Knowledge in Language Models Back to theTraining Data [Akyürek]arxiv.org/pdf/2205.11482
We Are All Software - Joscha Bach Machine Learning Street Talk 2024-08-21 | Dr. Joscha Bach introduces a surprising idea called "cyber animism" in his AGI-24 talk - the notion that nature might be full of self-organizing software agents, similar to the spirits in ancient belief systems. Bach suggests that consciousness could be a kind of software running on our brains, and wonders if similar "programs" might exist in plants or even entire ecosystems. MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. Joscha takes us on a tour de force through history, philosophy, and cutting-edge computer science, teasing us to rethink what we know about minds, machines, and the world around us. Joscha believes we should blur the lines between human, artificial, and natural intelligence, and argues that consciousness might be more widespread and interconnected than we ever thought possible.Dr. Joscha Bachhttps://x.com/PlinzThis is video 2/9 from our coverage of AGI-24 in Seattle agi-conf.org/2024 Watch the official MLST interview with Joscha which we did right after this talk on our Patreon now on early access - patreon.com/posts/joscha-bach-110199676 (you also get access to our private discord and biweekly calls)Pod: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Joscha-Bach---AGI24-Keynote-Cyberanimism-e2neeuiTOC:00:00:00 Introduction: AGI and Cyberanimism00:03:57 The Nature of Consciousness00:08:46 Aristotle's Concepts of Mind and Consciousness00:13:23 The Hard Problem of Consciousness00:16:17 Functional Definition of Consciousness00:20:24 Comparing LLMs and Human Consciousness00:26:52 Testing for Consciousness in AI Systems00:30:00 Animism and Software Agents in Nature00:37:02 Plant Consciousness and Ecosystem Intelligence00:40:36 The California Institute for Machine Consciousness00:44:52 Ethics of Conscious AI and Suffering00:46:29 Philosophical Perspectives on Consciousness00:49:55 Q&A: Formalisms for Conscious Systems00:53:27 Coherence, Self-Organization, and Compute ResourcesRefs:Aristotle's work on the soul and consciousnessRichard Dawkins' work on genes and evolutionGerald Edelman's concept of Neural DarwinismThomas Metzinger's book "Being No One"Yoshua Bengio's concept of the "consciousness prior"Stuart Hameroff's theories on microtubules and consciousnessChristof Koch's work on consciousnessDaniel Dennett's "Cartesian Theater" conceptGiulio Tononi's Integrated Information TheoryMike Levin's work on organismal intelligenceThe concept of animism in various culturesFreud's model of the mindBuddhist perspectives on consciousness and meditationThe Genesis creation narrative (for its metaphorical interpretation)California Institute for Machine Consciousness
The AI Bubble: Will It Burst, and What Comes After? Machine Learning Street Talk 2024-08-17 | Prof Gary Marcus revisited his keynote from AGI-21, noting that many of the issues he highlighted then are still relevant today despite significant advances in AI.This is part 1, we will be releasing an in-depth interview with Gary in the coming weeks. MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. Gary Marcus criticized current large language models (LLMs) and generative AI for their unreliability, tendency to hallucinate, and inability to truly understand concepts.Marcus argued that the AI field is experiencing diminishing returns with current approaches, particularly the "scaling hypothesis" that simply adding more data and compute will lead to AGI.He advocated for a hybrid approach to AI that combines deep learning with symbolic AI, emphasizing the need for systems with deeper conceptual understanding.Marcus highlighted the importance of developing AI with innate understanding of concepts like space, time, and causality.He expressed concern about the moral decline in Silicon Valley and the rush to deploy potentially harmful AI technologies without adequate safeguards.Marcus predicted a possible upcoming "AI winter" due to inflated valuations, lack of profitability, and overhyped promises in the industry.He stressed the need for better regulation of AI, including transparency in training data, full disclosure of testing, and independent auditing of AI systems.Marcus proposed the creation of national and global AI agencies to oversee the development and deployment of AI technologies.He concluded by emphasizing the importance of interdisciplinary collaboration, focusing on robust AI with deep understanding, and implementing smart, agile governance for AI and AGI.Pre-order Gary's new book here:Taming Silicon Valley: How We Can Ensure That AI Works for Usamzn.to/4fO46pYFilmed at the AGI-24 conference: agi-conf.org/2024Refs:Closed source vs open-source models slide ~24 minsFine-tune Llama 3.1 Ultra-Efficiently with Unsloth (Maxime Labonne/Liquid AI)huggingface.co/blog/mlabonne/sft-llama3TOC:00:00:00 Introduction00:02:34 Introduction by Ben G00:05:17 Gary Marcus begins talk00:07:38 Critiquing current state of AI00:12:21 Lack of progress on key AI challenges00:16:05 Continued reliability issues with AI00:19:54 Economic challenges for AI industry00:25:11 Need for hybrid AI approaches00:29:58 Moral decline in Silicon Valley00:34:59 Risks of current generative AI00:40:43 Need for AI regulation and governance00:49:21 Concluding thoughts00:54:38 Q&A: Cycles of AI hype and winters01:00:10 Predicting a potential AI winter01:02:46 Discussion on interdisciplinary approach01:05:46 Question on regulating AI01:07:27 Ben G's perspective on AI winter
Is ChatGPT an N-gram model on steroids? Machine Learning Street Talk 2024-08-15 | DeepMind machine learning scientist / MIT scholar Dr. Timothy Nguyen discusses his recent paper on understanding transformers through n-gram statistics. Nguyen explains his approach to analyzing transformer behavior using a kind of "template matching" (N-grams), providing insights into how these models process and predict language.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. Key points covered include:A method for describing transformer predictions using n-gram statistics without relying on internal mechanisms.The discovery of a technique to detect overfitting in large language models without using holdout sets.Observations on curriculum learning, showing how transformers progress from simpler to more complex rules during training.Discussion of distance measures used in the analysis, particularly the variational distance.Exploration of model sizes, training dynamics, and their impact on the results.We also touch on philosophical aspects of describing versus explaining AI behavior, and the challenges in understanding the abstractions formed by neural networks. Nguyen concludes by discussing potential future research directions, including attempts to convert descriptions of transformer behavior into explanations of internal mechanisms.Timothy Nguyen's earned his B.S. and Ph.D. in mathematics from Caltech and MIT, respectively. He held positions as Research Assistant Professor at the Simons Center for Geometry and Physics (2011-2014) and Visiting Assistant Professor at Michigan State University (2014-2017). During this time, his research expanded into high-energy physics, focusing on mathematical problems in quantum field theory. His work notably provided a simplified and corrected formulation of perturbative path integrals.Since 2017, Nguyen has been working in industry, applying his expertise to machine learning. He is currently at DeepMind, where he contributes to both fundamental research and practical applications of deep learning to solve real-world problems.Refs:The Cartesian Cafeyoutube.com/@TimothyNguyenUnderstanding Transformers via N-Gram Statisticsresearchgate.net/publication/382204056_Understanding_Transformers_via_N-Gram_StatisticsTOC00:00:00 Timothy Nguyen's background00:02:50 Paper overview: transformers and n-gram statistics00:04:55 Template matching and hash table approach00:08:55 Comparing templates to transformer predictions00:12:01 Describing vs explaining transformer behavior00:15:36 Detecting overfitting without holdout sets00:22:47 Curriculum learning in training00:26:32 Distance measures in analysis00:28:58 Model sizes and training dynamics00:30:39 Future research directions00:32:06 Conclusion and future topics
Jay Alammar on LLMs, RAG, and AI Engineering Machine Learning Street Talk 2024-08-11 | Jay Alammar, renowned AI educator and researcher at Cohere, discusses the latest developments in large language models (LLMs) and their applications in industry. Jay shares his expertise on retrieval augmented generation (RAG), semantic search, and the future of AI architectures.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. Cohere Command R model series: cohere.com/commandJay Alamaar:https://x.com/jayalammarBuy Jay's new book here!Hands-On Large Language Models: Language Understanding and Generationamzn.to/4fzOUghTOC:00:00:00 Introduction to Jay Alammar and AI Education00:01:47 Cohere's Approach to RAG and AI Re-ranking00:07:15 Implementing AI in Enterprise: Challenges and Solutions00:09:26 Jay's Role at Cohere and the Importance of Learning in Public00:15:16 The Evolution of AI in Industry: From Deep Learning to LLMs00:26:12 Expert Advice for Newcomers in Machine Learning00:32:39 The Power of Semantic Search and Embeddings in AI Systems00:37:59 Jay Alammar's Journey as an AI Educator and Visualizer00:43:36 Visual Learning in AI: Making Complex Concepts Accessible00:47:38 Strategies for Keeping Up with Rapid AI Advancements00:49:12 The Future of Transformer Models and AI Architectures00:51:40 Evolution of the Transformer: From 2017 to Present00:54:19 Preview of Jay's Upcoming Book on Large Language ModelsDisclaimer: This is the fourth video from our Cohere partnership. We were not told what to say in the interview, and didn't edit anything out from the interview. Note also that this combines several previously unpublished interviews from Jay into one, the earlier one at Tim's house was shot in Aug 2023, and the more recent one in Toronto in May 2024. Refs:The Illustrated Transformerjalammar.github.io/illustrated-transformerAttention Is All You Needarxiv.org/abs/1706.03762The Unreasonable Effectiveness of Recurrent Neural Networkshttp://karpathy.github.io/2015/05/21/rnn-effectivenessNeural Networks in 11 Lines of Codeiamtrask.github.io/2015/07/12/basic-python-networkUnderstanding LSTM Networks (Chris Olah's blog post)http://colah.github.io/posts/2015-08-Understanding-LSTMsLuis Serrano's YouTube Channelyoutube.com/channel/UCgBncpylJ1kiVaPyP-PZauQSentence-BERT: Sentence Embeddings using Siamese BERT-Networksarxiv.org/abs/1908.10084GPT (Generative Pre-trained Transformer) modelsjalammar.github.io/illustrated-gpt2openai.com/research/gpt-4BERT (Bidirectional Encoder Representations from Transformers)jalammar.github.io/illustrated-bertarxiv.org/abs/1810.04805RoPE (Rotary Positional Encoding)arxiv.org/abs/2104.09864 (Linked paper discussing rotary embeddings)Grouped Query Attentionarxiv.org/pdf/2305.13245RLHF (Reinforcement Learning from Human Feedback)openai.com/research/learning-from-human-preferencesarxiv.org/abs/1706.03741DPO (Direct Preference Optimization)arxiv.org/abs/2305.18290
Daniel Cahn - Slingshot AI (AI Therapy) Machine Learning Street Talk 2024-08-08 | Daniel Cahn, co-founder of Slingshot AI, on the potential of AI in therapy and modelling the mind with AI. Why is anxiety and depression affecting a large population? To what extent are these real categories? Why is the mental health getting worse? How often do you want an AI to agree with you? What are the ethics of persuasive AI? The answers to these questions will surprise you.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. Daniel Cahn (who is also hiring ML engineers by the way!)https://x.com/thecahnartist?lang=enlinkedin.com/in/cahndthinkingmachinespodcast.comTOC:00:00:00 Intro00:01:56 Therapy effectiveness vs drugs and societal implications00:04:02 Mental health categories: Iatrogenesis and social constructs00:10:19 Psychiatric treatment models and cognitive perspectives00:13:30 AI design and human-like interactions: Intentionality debates00:20:04 AI in therapy: Ethics, anthropomorphism, and loneliness mitigation00:28:13 Therapy efficacy: Neuroplasticity, suffering, and AI placebos00:33:29 AI's impact on human agency and cognitive modeling00:41:17 Social media's effects on brain structure and behavior00:50:46 AI ethics: Altering values and free will considerations01:00:00 Work value perception and personal identity formation01:13:37 Free will, agency, and mutable personal identity in therapy01:24:27 AI in healthcare: Challenges, ethics, and therapy improvements01:53:25 AI development: Societal impacts and cultural implicationsPod version: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Can-AI-therapy-be-more-effective-than-drugs-e2mvhd0Refs:Bad Therapy: Why the Kids Aren't Growing Up (Abigail Shrier)amzn.to/3WSLga4Irreversible Damage: Teenage Girls and the Transgender Craze (Shrier)amzn.to/4d9kpfx (note that this book will be offensive to trans folks, forewarned - Tim hasn't read it/isn't taking a position)The Coddling of the American Mind (Haidt)amzn.to/3WMtPaSThe Anxious Generation (Haidt)amzn.to/3SB9MJULost Connections (Johan Hari)amzn.to/3SC7471(Note the book cover of "Hillbilly Elegy" was a mistake, I meant to link to "The Minds of Billy Milligan") i.e. a book about multiple personality disorderamzn.to/4dUHZg7Noise (Kahneman)amzn.to/4ftMgsiDiagnostic and Statistical Manual of Mental Disorders (DSM-5-TR)psychiatry.org/psychiatrists/practice/dsmBiopsychosocial modelen.wikipedia.org/wiki/Biopsychosocial_modelThe serotonin theory of depression: how the media got it all wrongpharmaceutical-journal.com/article/feature/the-serotonin-theory-of-depression-how-the-media-got-it-all-wrongThe ELIZA effecten.wikipedia.org/wiki/ELIZA_effectHow to Win Friends and Influence Peopleamazon.co.uk/How-Win-Friends-Influence-People/dp/0091906814Evidence of Human-Level Bonds Established With a Digital Conversational Agent: Cross-sectional, Retrospective Observational Study (JMIR)formative.jmir.org/2021/5/e27868COMPUTING MACHINERY AND INTELLIGENCE (Turing - imitation game)cs.ox.ac.uk/activities/ieg/e-library/sources/t_article.pdfDecision transformersarxiv.org/abs/2106.01345Loneliness and suicide mitigation for students using GPT3-enabled chatbotsnature.com/articles/s44184-023-00047-6Little Treatments, Big Effects (Jessica Schleider)amzn.to/4deN7vhWhy Greatness Cannot Be Planned (Stanley and Lehman) [Possibly the best book ever written]amzn.to/3yxykNfMachine Love (Lehman)arxiv.org/abs/2302.09248Infinite gamesamzn.to/3LYF3TqOpenAI model specopenai.com/index/introducing-the-model-specA Brief History of Intelligence (Bennett - interview dropping soon)amzn.to/3WTWs5ZBullshit Jobs: A Theoryamzn.to/46ENSeOJohn Carmack take - https://x.com/ID_AA_Carmack/status/1796622337963389412Determined (Robert M Sapolsky)amzn.to/4dc8KN3So You've Been Publicly Shamed (Jon Ronson)amzn.to/3WDLzE9The Examined Life (Stephen Grosz)amzn.to/3SF6s0HThinking, Fast and Slow (Daniel Kahneman)amzn.to/3WzEZyxDecision Transformer: Reinforcement Learning via Sequence Modelingarxiv.org/pdf/2106.01345Ender's Gameamzn.to/3yxXNGqMagic Pill: The Extraordinary Benefits and Disturbing Risks of the New Weight Loss Drugsamzn.to/4ccrc6QThinking Machines (Daniel’s podcast!)thinkingmachinespodcast.comThe Upside of Stress (Daniel referenced this on a part of the conversation which we edited out) amzn.to/3LYMTww
Do you think that ChatGPT can reason? Machine Learning Street Talk 2024-07-29 | Prof. Subbarao Kambhampati argues that while LLMs are impressive and useful tools, especially for creative tasks, they have fundamental limitations in logical reasoning and cannot provide guarantees about the correctness of their outputs. He advocates for hybrid approaches that combine LLMs with external verification systems.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. This is 2/13 of our #ICML2024 seriesTOC[00:00:00] Intro[00:02:06] Bio[00:03:02] LLMs are n-gram models on steroids[00:07:26] Is natural language a formal language?[00:08:34] Natural language is formal?[00:11:01] Do LLMs reason?[00:19:13] Definition of reasoning[00:31:40] Creativity in reasoning[00:50:27] Chollet's ARC challenge[01:01:31] Can we reason without verification?[01:10:00] LLMs cant solve some tasks[01:19:07] LLM Modulo framework[01:29:26] Future trends of architecture[01:34:48] Future research directionsPod: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Prof--Subbarao-Kambhampati---LLMs-dont-reason--they-memorize-ICML2024-213-e2mjcseSubbarao Kambhampati:https://x.com/rao2zInterviewer: Dr. Tim ScarfeRefs:Can LLMs Really Reason and Plan?cacm.acm.org/blogcacm/can-llms-really-reason-and-planOn the Planning Abilities of Large Language Models : A Critical Investigationarxiv.org/pdf/2305.15771Chain of Thoughtlessness? An Analysis of CoT in Planningarxiv.org/pdf/2405.04776On the Self-Verification Limitations of Large Language Models on Reasoning and Planning Tasksarxiv.org/pdf/2402.08115LLMs Can't Plan, But Can Help Planning in LLM-Modulo Frameworksarxiv.org/pdf/2402.01817Embers of Autoregression: Understanding Large LanguageModels Through the Problem They are Trained to Solvearxiv.org/pdf/2309.13638arxiv.org/abs/2402.04210"Task Success" is not EnoughFaith and Fate: Limits of Transformers on Compositionality "finetuning multiplication with four digit numbers" (added after pub)arxiv.org/pdf/2305.18654Partition function (number theory) (Srinivasa Ramanujan and G.H. Hardy's work)en.wikipedia.org/wiki/Partition_function_(number_theory)Poincaré conjectureen.wikipedia.org/wiki/Poincar%C3%A9_conjectureGödel's incompleteness theoremsen.wikipedia.org/wiki/G%C3%B6del%27s_incompleteness_theoremsROT13 (Rotate13, "rotate by 13 places")en.wikipedia.org/wiki/ROT13A Mathematical Theory of Communication (C. E. SHANNON)https://people.math.harvard.edu/~ctm/home/text/others/shannon/entropy/entropy.pdfSparks of AGIarxiv.org/abs/2303.12712Kambhampati thesis on speech recognition (1983)https://rakaposhi.eas.asu.edu/rao-btech-thesis.pdfPlanBench: An Extensible Benchmark for Evaluating Large Language Models on Planning and Reasoning about Changearxiv.org/abs/2206.10498Explainable human-AI interactionlink.springer.com/book/10.1007/978-3-031-03767-2Tree of Thoughtsarxiv.org/abs/2305.10601On the Measure of Intelligence (ARC Challenge)arxiv.org/abs/1911.01547Getting 50% (SoTA) on ARC-AGI with GPT-4o (Ryan Greenblatt ARC solution)redwoodresearch.substack.com/p/getting-50-sota-on-arc-agi-with-gptPROGRAMS WITH COMMON SENSE (John McCarthy) - "AI should be an advice taker program"https://www.cs.cornell.edu/selman/cs672/readings/mccarthy-upd.pdfOriginal chain of thought paperarxiv.org/abs/2201.11903ICAPS 2024 Keynote: Dale Schuurmans on "Computing and Planning with Large Generative Models" (COT)youtube.com/watch?v=YnMqbpdHcaYThe Hardware Lottery (Hooker)arxiv.org/abs/2009.06489A Path Towards Autonomous Machine Intelligence (JEPA/LeCun)openreview.net/pdf?id=BZ5a1r-kVsfAlphaGeometrynature.com/articles/s41586-023-06747-5FunSearchnature.com/articles/s41586-023-06924-6Emergent Abilities of Large Language Modelsarxiv.org/abs/2206.07682Language models are not naysayers (Negation in LLMs)arxiv.org/abs/2306.08189The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"arxiv.org/abs/2309.12288Embracing negative resultsopenreview.net/forum?id=3RXAiU7sss
The threat of existential risk from AI Machine Learning Street Talk 2024-07-28 | How seriously should governments take the threat of existential risk from AI, given the lack of consensus among researchers? On the one hand, existential risks (x-risks) are necessarily somewhat speculative: by the time there is concrete evidence, it may be too late. On the other hand, governments must prioritize — after all, they don’t worry too much about x-risk from alien invasions.MLST is sponsored by Brave:The Brave Search API covers over 20 billion webpages, built from scratch without Big Tech biases or the recent extortionate price hikes on search API access. Perfect for AI model training and retrieval augmentated generation. Try it now - get 2,000 free queries monthly at http://brave.com/api. This is part of our ICML 2024 series (1/13)Sayash Kapoor is a computer science Ph.D. candidate at Princeton University's Center for Information Technology Policy. His research focuses on the societal impact of AI. Kapoor has previously worked on AI in both industry and academia, with experience at Facebook, Columbia University, and EPFL Switzerland. He is a recipient of a best paper award at ACM FAccT and an impact recognition award at ACM CSCW. Notably, Kapoor was included in TIME's inaugural list of the 100 most influential people in AI.Sayash Kapoorhttps://x.com/sayashkhttps://www.cs.princeton.edu/~sayashk/Arvind Narayanan (other half of the AI Snake Oil duo)https://x.com/random_walkerAI existential risk probabilities are too unreliable to inform policyaisnakeoil.com/p/ai-existential-risk-probabilitiesPre-order AI Snake Oil Bookamzn.to/4fq2HGbAI Snake Oil blogaisnakeoil.comAI Agents That Matterarxiv.org/abs/2407.01502Shortcut learning in deep neural networkssemanticscholar.org/paper/Shortcut-learning-in-deep-neural-networks-Geirhos-Jacobsen/1b04936c2599e59b120f743fbb30df2eed3fd78277% Of Employees Report AI Has Increased Workloads And Hampered Productivity, Study Findsforbes.com/sites/bryanrobinson/2024/07/23/employees-report-ai-increased-workloadPod version: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Sayash-Kapoor---How-seriously-should-we-take-AI-X-risk--ICML-113-e2mhuocTOC:00:00:00 Intro00:01:57 How seriously should we take Xrisk threat?00:02:55 Risk too unrealiable to inform policy00:10:20 Overinflated risks00:12:05 Perils of utility maximisation00:13:55 Scaling vs airplane speeds00:17:31 Shift to smaller models?00:19:08 Commercial LLM ecosystem00:22:10 Synthetic data00:24:09 Is AI complexifying our jobs?00:25:50 Does ChatGPT make us dumber or smarter?00:26:55 Are AI Agents overhyped?00:28:12 Simple vs complex baselines00:30:00 Cost tradeoff in agent design00:32:30 Model eval vs downstream perf00:36:49 Shortcuts in metrics00:40:09 Standardisation of agent evals00:41:21 Humans in the loop00:43:54 Levels of agent generality00:47:25 ARC challenge
Science should be self correcting - Subbarao Kambhampati Machine Learning Street Talk 2024-07-24 | ...
Why US AI Act Compute Thresholds Are Misguided... Machine Learning Street Talk 2024-07-18 | Sara Hooker is VP of Research at Cohere and leader of Cohere for AI. We discuss her recent paper critiquing the use of compute thresholds, measured in FLOPs (floating point operations), as an AI governance strategy. We explore why this approach, recently adopted in both US and EU AI policies, may be problematic and oversimplified. Sara explains the limitations of using raw computational power as a measure of AI capability or risk, and discusses the complex relationship between compute, data, and model architecture.Equally important, we go into Sara's work on "The AI Language Gap." This research highlights the challenges and inequalities in developing AI systems that work across multiple languages. Sara discusses how current AI models, predominantly trained on English and a handful of high-resource languages, fail to serve the linguistic diversity of our global population. We explore the technical, ethical, and societal implications of this gap, and discuss potential solutions for creating more inclusive and representative AI systems.We broadly discuss the relationship between language, culture, and AI capabilities, as well as the ethical considerations in AI development and deployment.Pod version: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Sara-Hooker---Why-US-AI-Act-Compute-Thresholds-Are-Misguided-e2m6qm4TOC:[00:00:00] Intro[00:02:12] FLOPS paper[00:26:42] Hardware lottery[00:30:22] The Language gap[00:33:25] Safety[00:38:31] Emergent[00:41:23] Creativity[00:43:40] Long tail[00:44:26] LLMs and society[00:45:36] Model bias[00:48:51] Language and capabilities[00:52:27] Ethical frameworks and RLHFSara Hookersarahooker.melinkedin.com/in/sararosehookerscholar.google.com/citations?user=2xy6h3sAAAAJ&hl=enhttps://x.com/sarahookrInterviewer: Tim ScarfeRefsThe AI Language gapcohere.com/research/papers/the-AI-language-gap.pdfOn the Limitations of Compute Thresholds as a Governance Strategy.arxiv.org/pdf/2407.05694v1The Multilingual Alignment Prism: Aligning Global and Local Preferences to Reduce Harmarxiv.org/pdf/2406.18682Cohere Ayacohere.com/research/ayaRLHF Can Speak Many Languages: Unlocking Multilingual Preference Optimization for LLMsarxiv.org/pdf/2407.02552Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMsarxiv.org/pdf/2402.14740Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligencewhitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligenceEU AI Acthttps://www.europarl.europa.eu/doceo/document/TA-9-2024-0138_EN.pdfThe bitter lessonhttp://www.incompleteideas.net/IncIdeas/BitterLesson.htmlNeel Nanda interviewyoutube.com/watch?v=_Ygf0GnlwmYScaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnethttps://transformer-circuits.pub/2024/scaling-monosemanticity/Chollet's ARC challengegithub.com/fchollet/ARC-AGIRyan Greenblatt on ARCyoutube.com/watch?v=z9j3wB1RRGADisclaimer: This is the third video from our Cohere partnership. We were not told what to say in the interview, and didn't edit anything out from the interview.
There are monsters in your LLM. Machine Learning Street Talk 2024-07-14 | Murray Shanahan is a professor of Cognitive Robotics at Imperial College London and a senior research scientist at DeepMind. He challenges our assumptions about AI consciousness and urges us to rethink how we talk about machine intelligence. We explore the dangers of anthropomorphizing AI, the limitations of current language in describing AI capabilities, and the fascinating intersection of philosophy and artificial intelligence.Show notes and full references: docs.google.com/document/d/1ICtBI574W-xGi8Z2ZtUNeKWiOiGZ_DRsp9EnyYAISws/edit?usp=sharing Prof Murray Shanahan:doc.ic.ac.uk/~mpsha (look at his selected publications)scholar.google.co.uk/citations?user=00bnGpAAAAAJ&hl=enen.wikipedia.org/wiki/Murray_Shanahan https://x.com/mpshanahan Pod: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Prof--Murray-Shanahan---Machines-Dont-Think-Like-Us-e2m1499Interviewer: Dr. Tim ScarfeRefs (links in the Google doc linked above):Role play with large language modelsWaluigi effect"Conscious Exotica" - Paper by Murray Shanahan (2016)"Simulators" - Article by Janis from LessWrong"Embodiment and the Inner Life" - Book by Murray Shanahan (2010)"The Technological Singularity" - Book by Murray Shanahan (2015)"Simulacra as Conscious Exotica" - Paper by Murray Shanahan (newer paper of the original focussed on LLMs)A recent paper by Anthropic on using autoencoders to find features in language models (referring to the "Scaling Monosemanticity" paper)Work by Peter Godfrey-Smith on octopus consciousness"Metaphors We Live By" - Book by George Lakoff (1980s)Work by Aaron Sloman on the concept of "space of possible minds" (1984 article mentioned)Wittgenstein's "Philosophical Investigations" (posthumously published)Daniel Dennett's work on the "intentional stance"Alan Turing's original paper on the Turing Test (1950)Thomas Nagel's paper "What is it like to be a bat?" (1974)John Searle's Chinese Room Argument (mentioned but not detailed)Work by Richard Evans on tackling reasoning problemsClaude Shannon's quote on knowledge and control"Are We Bodies or Souls?" - Book by Richard SwinburneReference to work by Ethan Perez and others at Anthropic on potential deceptive behavior in language modelsReference to a paper by Murray Shanahan and Antonia Creswell on the "selection inference framework"Mention of work by Francois Chollet, particularly the ARC (Abstraction and Reasoning Corpus) challengeReference to Elizabeth Spelke's work on core knowledge in infantsMention of Karl Friston's work on planning as inference (active inference)The film "Ex Machina" - Murray Shanahan was the scientific advisor"The Waluigi Effect"Anthropic's constitutional AI approachLoom system by Lara Reynolds and Kyle McDonald for visualizing conversation treesDeepMind's AlphaGo (mentioned multiple times as an example)Mention of the "Golden Gate Claude" experimentReference to an interview Tim Scarfe conducted with University of Toronto students about self-attention controllability theoremMention of an interview with Irina RishReference to an interview Tim Scarfe conducted with Daniel DennettReference to an interview with Maria Santa CaterinaMention of an interview with Philip GoffNick Chater and Martin Christianson's book ("The Language Game: How Improvisation Created Language and Changed the World")Peter Singer's work from 1975 on ascribing moral status to conscious beingsDemis Hassabis' discussion on the "ladder of creativity"Reference to B.F. Skinner and behaviorismTOC:00:00:00 Intro00:05:49 Simulators00:11:04 The 20 questions game and simulacra stickiness00:18:50 Murray's experience with Claude 300:30:04 RLHF00:32:41 Anthropic Golden Gate Bridge00:37:05 Agency00:41:05 Embodiment and knowledge acquisition00:57:51 ARC01:03:31 The conscious stance01:13:58 Space of possible minds01:11:05 part 2: Wittgenstein private language argument / subjectivity01:29:58 Conscious exotica01:33:23 Dennett and intentional stance01:40:58 Anthropomorphisation01:46:47 Reasoning01:53:56 Turing test02:04:41 Nagel's bat02:08:08 Mark Bishop and Idealism/CRA02:09:32 Panpsychism
Solving Chollets ARC-AGI with GPT4o Machine Learning Street Talk 2024-07-06 | Ryan Greenblatt from Redwood Research recently published "Getting 50% on ARC-AGI with GPT-4.0," where he used GPT4o to reach a state-of-the-art accuracy on Francois Chollet's ARC Challenge by generating many Python programs. Sponsor:Sign up to Kalshi here kalshi.onelink.me/1r91/mlst -- the first 500 traders who deposit $100 will get a free $20 credit! Important disclaimer - In case it's not obvious - this is basically gambling and a *high risk* activity - only trade what you can afford to lose.We discuss:- Ryan's unique approach to solving the ARC Challenge and achieving impressive results.- The strengths and weaknesses of current AI models.- How AI and humans differ in learning and reasoning.- Combining various techniques to create smarter AI systems.- The potential risks and future advancements in AI, including the idea of agentic AI.https://x.com/RyanPGreenblattredwoodresearch.orgTOC00:00:00 Intro00:01:38 Prelude on goals in LLMs00:02:42 Ryan intro00:03:11 Ryan's ARC Challenge Approach00:38:15 Language models, reasoning and agency01:14:14 Timelines on superintelligence01:27:05 Growth of superintelligence02:06:41 Reflections on ARC02:11:49 Why wouldn't AI knowledge be subjectiveHost: Dr. Tim ScarfePod: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Ryan-Greenblatt---Solving-ARC-with-GPT4o-e2lnplqRefs:Getting 50% (SoTA) on ARC-AGI with GPT-4o [Ryan Greenblatt]redwoodresearch.substack.com/p/getting-50-sota-on-arc-agi-with-gptOn the Measure of Intelligence [Chollet]arxiv.org/abs/1911.01547Connectionism and Cognitive Architecture: A Critical Analysis [Jerry A. Fodor and Zenon W. Pylyshyn]https://ruccs.rutgers.edu/images/personal-zenon-pylyshyn/proseminars/Proseminar13/ConnectionistArchitecture.pdfSoftware 2.0 [Andrej Karpathy]karpathy.medium.com/software-2-0-a64152b37c35Why Greatness Cannot Be Planned: The Myth of the Objective [Kenneth Stanley]amzn.to/3Wfy2E0Biographical account of Terence Tao’s mathematical development. [M.A.(KEN) CLEMENTS]gwern.net/doc/iq/high/smpy/1984-clements.pdfModel Evaluation and Threat Research (METR)metr.orgWhy Tool AIs Want to Be Agent AIsgwern.net/tool-aiSimulators - Januslesswrong.com/posts/vJFdjigzmcXMhNTsx/simulatorsAI Control: Improving Safety Despite Intentional Subversionlesswrong.com/posts/d9FJHawgkiMSPjagR/ai-control-improving-safety-despite-intentional-subversionarxiv.org/abs/2312.06942What a Compute-Centric Framework Says About Takeoff Speedsopenphilanthropy.org/research/what-a-compute-centric-framework-says-about-takeoff-speedsGlobal GDP over the long runourworldindata.org/grapher/global-gdp-over-the-long-run?yScale=logSafety Cases: How to Justify the Safety of Advanced AI Systemsarxiv.org/abs/2403.10462The Danger of a “Safety Case"http://sunnyday.mit.edu/The-Danger-of-a-Safety-Case.pdfThe Future Of Work Looks Like A UPS Truck (~02:15:50)npr.org/sections/money/2014/05/02/308640135/episode-536-the-future-of-work-looks-like-a-ups-truckSWE-benchswebench.comUsing DeepSpeed and Megatron to Train Megatron-Turing NLG530B, A Large-Scale Generative Language Modelarxiv.org/pdf/2201.11990 Algorithmic Progress in Language Models epochai.org/blog/algorithmic-progress-in-language-models
How Cohere will improve AI Reasoning this year Machine Learning Street Talk 2024-06-29 | Aidan Gomez, CEO of Cohere, reveals how they're tackling AI hallucinations and improving reasoning abilities. He also explains why Cohere doesn't use any output from GPT-4 for training their models.Aidan shares his personal insights into the world of AI and LLMs and Cohere's unique approach to solving real-world business problems, and how their models are set apart from the competition. Aidan reveals how they are making major strides in AI technology, discussing everything from last mile customer engineering to the robustness of prompts and future architectures. He also touches on the broader implications of AI for society, including potential risks and the role of regulation. He discusses Cohere's guiding principles and the health the of startup scene. With a particular focus on enterprise applications. Aidan provides a rare look into the internal workings of Cohere and their vision for driving productivity and innovation.cohere.comhttps://x.com/aidangomezCheck out Cohere's amazing new Command R* models herecohere.com/commandPod: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Aiden-Gomez---CEO-of-Cohere-AIs-Inner-Monologue--Crucial-for-Reasoning-e2lf8fgTOC:00:00:00 Intro00:01:48 Guiding principles of Cohere00:02:31 Last mile / customer engineering00:04:25 Prompt brittleness00:06:14 Robustness and "delving"00:10:12 Command R models and catch up00:12:32 Are LLMs saturating / specialisation00:16:11 Intelligence00:21:28 Predictive architectures, data vs inductive priors00:25:55 Agentic systems00:28:11 Differentiation00:33:35 X-Risk / Bostrom00:39:30 Changing relationship with technology00:45:08 Policy00:49:01 Startup scene00:52:44 Biggest mistake?00:53:50 Management style00:56:38 Culture in different Cohere offices?Disclaimer: This is the second video from our Cohere partnership. We were not told what to say in the interview, and didn't edit anything out from the interview.
Could the universe be conscious? Machine Learning Street Talk 2024-06-21 | Philip Goff believes that everything, even tiny particles like electrons, has a little bit of consciousness. This idea is called panpsychism. He explains that this might help us understand why we have feelings and thoughts.Philip discuss another idea called cosmopsychism, which is a theory that suggests the entire universe is a single conscious entity. Instead of individual minds (like human minds) being separate and independent, they are seen as parts of the universe's larger, unified consciousness. In simpler terms, it means that the universe itself has a mind, and our individual consciousnesses are just small parts of this greater, universal mind.Philip thinks science can't fully explain what it's like to feel things, like the color red or the taste of chocolate. He says we need to include consciousness in our science to understand everything better.We edited out discussion of the Twitter argument with Sabine on this video, but you can find it on Twitter here - https://x.com/MLStreetTalk/status/1804825098207052061https://x.com/Philip_GoffWhy? The Purpose of the Universe (Prof Philip Goff)amzn.to/4cbYHqLGalileo's Error: Foundations for a New Science of Consciousnessamzn.to/4eqA5MiConsciousness and Fundamental Realityamzn.to/3VDd8wXIs Consciousness Everywhere?amzn.to/3xAiC3mTOC00:00:00 Introduction to Consciousness00:02:52 Panpsychism Explained00:04:43 Cosmopsychism and Panagentialism00:08:44 Exploring Agency and Purpose00:22:51 Physicalism vs. Panpsychism00:28:03 Critique of Philosophical Views on Consciousness00:30:51 Frank Jackson's Knowledge Argument00:33:50 Galileo and the Foundations of Physical Science00:36:43 Philosophical and Scientific Integration00:43:57 MindChat and Constructive Disagreement00:46:29 Book signing and Final ThoughtsWe used a clip of Sam Harris from the Know Thyself podcast - youtube.com/watch?v=gqA-ZRpl1jQ
Chollets ARC Challenge + Current Winners Machine Learning Street Talk 2024-06-18 | The ARC Challenge, created by Francois Chollet, tests how well AI systems can generalize from a few examples in a grid-based intelligence test. We interview the current winners of the ARC Challenge—Jack Cole, Mohammed Osman and their collaborator Michael Hodel. They discuss how they tackled ARC (Abstraction and Reasoning Corpus) using language models. We also discuss the new "50%" public set approach announced today from Redwood Research (Ryan Greenblatt).Jack and Mohammed explain their winning approach, which involves fine-tuning a language model on a large, specifically-generated dataset and then doing additional fine-tuning at test-time, a technique known in this context as "active inference". They use various strategies to represent the data for the language model and believe that with further improvements, the accuracy could reach above 50%. Michael talks about his work on generating new ARC-like tasks to help train the models. They also debate whether their methods stay true to the "spirit" of Chollet's measure of intelligence. Despite some concerns, they agree that their solutions are promising and adaptable for other similar problems. Note:Jack's team is still the current official winner at 33% on the private set. Ryan's entry is not on the private leaderboard or eligible.Chollet invented ARC in 2019 (not 2017 as stated)"Ryan's entry is not a new state of the art. We don't know exactly how well it does since it was only evaluated on 100 tasks from the evaluation set and does 50% on those, reportedly. Meanwhile Jacks team i.e. MindsAI's solution does 54% on the entire eval set and it is seemingly possible to do 60-70% with an ensemble"Jack Cole:https://x.com/Jcole75Colehttps://lab42.global/community-interview-jack-cole/Mohamed Osman:Mohamed is looking to do a PhD in AI/ML, can you help him? Email: mothman198@outlook.comlinkedin.com/in/mohamedosman1905Michael Hodel:arxiv.org/pdf/2404.07353v1linkedin.com/in/michael-hodelhttps://x.com/bayesilicongithub.com/michaelhodelGetting 50% (SoTA) on ARC-AGI with GPT-4o - Ryan Greenblattredwoodresearch.substack.com/p/getting-50-sota-on-arc-agi-with-gptNeural networks for abstraction and reasoning: Towards broad generalization in machines [Mikel Bober-Irizar, Soumya Banerjee]arxiv.org/pdf/2402.03507Measure of intelligence:arxiv.org/abs/1911.01547I think the audio levelling might be a bit off on this for the intro especially, I fixed it on the audio podcast version - sorry if it's annoying. Pod version: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/New-50-ARC-result-and-current-winners-interviewed-e2l1prlTOC (autogenerated):00:00:00 Introduction00:03:00 Francois Chollet's Intelligence Concept00:08:00 Human Collaboration00:15:00 ARC Tasks and Symbolic AI00:27:00 Evaluation Techniques00:35:23 (Main Interview) Competitors and Approaches00:40:00 Meta Learning Challenges00:48:00 System 1 vs System 201:00:00 Inductive Priors and Symbols01:18:00 Methodologies Comparison01:25:00 Training Data Size Impact01:35:00 Generalization Issues01:47:00 Techniques for AI Applications01:56:00 Model Efficiency and Scalability02:10:00 Task Specificity and Generalization02:13:00 Summary
ARC challenge - where does the reasoning happen? #artificialintelligence  Machine Learning Street Talk 2024-06-18 | ...
Cohere is not an AGI company - Nick Frosst (co-founder) Machine Learning Street Talk 2024-06-16 | ...
Is AGI Just a Fantasy? Machine Learning Street Talk 2024-06-15 | Nick Frosst, the co-founder of Cohere, on the future of LLMs, and AGI. Learn how Cohere is solving real problems for business with their new AI models.Nick talks about his journey at Google Brain, working with AI legends like Geoff Hinton, and the amazing things his company, Cohere, is doing. From creating the must useful language models for businesses to making tools for developers, Nick shares a lot of interesting insights. He even talks about his band, Good Kid! Nick said that RAG is one of the best features of Cohere's new Command R* models. We are about to release a deep-dive on RAG with Patrick Lewis from Cohere, keep an eye out for that - he explains why their models are specifically optimised for RAG use cases. Learn more about Cohere Command R models here:cohere.com/commandgithub.com/cohere-ai/cohere-toolkitNick's band Good Kid:goodkidofficial.comNick on Twitter:https://x.com/nickfrosst00:00:00 Intro00:01:55 Backstory of Cohere00:02:31 Hinton00:02:54 Nick's band00:03:11 How is Cohere differentiated?00:03:44 Not an AGI company00:06:00 Command+R00:06:41 Standout feature: RAG00:09:07 How is RAG changing the way we build apps?00:09:44 Build day00:10:25 Building robust applications00:11:59 RAG evolution00:14:45 Unsupervised RAG00:16:30 Agents and divergence00:18:27 Agency00:22:19 Are LLMs general?00:24:48 Benchmarks00:27:07 Would Cohere verticalize?00:27:43 RAG vs long context00:29:20 Tech hasn't landed yet?00:31:36 Are LLMs saturating?00:35:50 Cohere's data acquisition pipeline00:36:34 SOTA chasing vs fairness00:37:21 Fall of the data scientist00:40:37 Final calloutsDisclaimer: This is the first video from our Cohere partnership. We were not told what to say in the interview, and didn't edit anything out from the interview.
Consciousness in LLMs Machine Learning Street Talk 2024-06-09 | patreon.com/posts/prof-murray-105678581 watch now on early access!
Prof. Mark Solms on philosophical clarity (on Patreon early-access now). Machine Learning Street Talk 2024-06-09 | ...
Prior knowledge in ChatGPT Machine Learning Street Talk 2024-05-27 | patreon.com/posts/elliot-murphy-105019438 watch early here (Elliot Murphy show)
Mapping GPT revealed something strange... Machine Learning Street Talk 2024-05-24 | These two scientists have mapped out the insides or “reachable space” of a language model using control theory, what they discovered was extremely surprising. Please support us on Patreon to get access to the private Discord server, bi-weekly calls, early access and ad-free listening.patreon.com/mlstAman Bhargava from Caltech and Cameron Witkowski from the University of Toronto to discuss their groundbreaking paper, “What’s the Magic Word? A Control Theory of LLM Prompting.” (the main theorem on self-attention controllability was developed in collaboration with Dr. Shi-Zhuo Looi from Caltech).They frame LLM systems as discrete stochastic dynamical systems. This means they look at LLMs in a structured way, similar to how we analyze control systems in engineering. They explore the “reachable set” of outputs for an LLM. Essentially, this is the range of possible outputs the model can generate from a given starting point when influenced by different prompts. The research highlights that prompt engineering, or optimizing the input tokens, can significantly influence LLM outputs. They show that even short prompts can drastically alter the likelihood of specific outputs. Aman and Cameron’s work might be a boon for understanding and improving LLMs. They suggest that a deeper exploration of control theory concepts could lead to more reliable and capable language models.We dropped an additional, more technical video on the research on our Twitter account here: https://x.com/MLStreetTalk/status/1795093759471890606Pod version with no music/SFX: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Whats-the-Magic-Word--A-Control-Theory-of-LLM-Prompting-e2khs2tAdditional 20 minutes of unreleased footage on our Patreon here: patreon.com/posts/whats-magic-word-104922629What's the Magic Word? A Control Theory of LLM Prompting (Aman Bhargava, Cameron Witkowski, Manav Shah, Matt Thomson)arxiv.org/abs/2310.04444LLM Control Theory Seminar (April 2024)youtube.com/watch?v=9QtS9sVBFM0Society for the pursuit of AGI (Cameron founded it)agisociety.mydurable.comRoger Federer demohttp://conway.languagegame.io/inferenceNeural Cellular Automata, Active Inference, and the Mystery of Biological Computation (Aman)aman-bhargava.com/ai/neuro/neuromorphic/2024/03/25/nca-do-active-inference.html Aman and Cameron also want to thank Dr. Shi-Zhuo Looi and Prof. Matt Thomson from from Caltech for help and advice on their research. (https://thomsonlab.caltech.edu/ and https://pma.caltech.edu/people/looi-shi-zhuo)https://x.com/ABhargava2000https://x.com/witkowski_camTOC:00:00:00 - Main Intro00:06:25 - Bios00:07:50 - Control Theory and Governors00:09:37 - LLM Control Theory00:17:17 - Federer Game00:19:49 - Building LLM Controllers00:20:56 - Priors in LLMs00:28:44 - Manipulating LLMs00:34:11 - Adversarial Examples and Robustification00:36:54 - Model vs Software00:39:12 - Experiments in the Paper00:44:36 - Language as an Interstate Freeway00:46:41 - Collective Intelligence00:58:54 - Biomimetic Intelligence01:03:37 - Society for the Pursuit of AGI01:05:47 - ICLR Rejection
CAN MACHINES REPLACE US? (AI vs Humanity) Machine Learning Street Talk 2024-05-06 | Maria Santacaterina, with her background in the humanities, brings a critical perspective on the current state and future implications of AI technology, its impact on society, and the nature of human intelligence and creativity. She emphasizes that despite technological advancements, AI lacks fundamental human traits such as consciousness, empathy, intuition, and the ability to engage in genuine creative processes. Maria argues that AI, at its core, processes data but does not have the capability to understand or generate new, intrinsic meaning or ideas as humans do.Throughout the conversation, Maria highlights her concern about the overreliance on AI in critical sectors such as healthcare, the justice system, and business. She stresses that while AI can serve as a tool, it should not replace human judgment and decision-making. Maria points out that AI systems often operate on past data, which may lead to outdated or incorrect decisions if not carefully managed.The discussion also touches upon the concept of "adaptive resilience", which Maria describes in her book. She explains adaptive resilience as the capacity for individuals and enterprises to evolve and thrive amidst challenges by leveraging technology responsibly, without undermining human values and capabilities.A significant portion of the conversation focussed on ethical considerations surrounding AI. Tim and Maria agree that there's a pressing need for strong governance and ethical frameworks to guide AI development and deployment. They discuss how AI, without proper ethical considerations, risks exacerbating issues like privacy invasion, misinformation, and unintended discrimination.Maria is skeptical about claims of achieving Artificial General Intelligence (AGI) or a technological singularity where machines surpass human intelligence in all aspects. She argues that such scenarios neglect the complex, dynamic nature of human intelligence and consciousness, which cannot be fully replicated or replaced by machines.Tim and Maria discuss the importance of keeping human agency and creativity at the forefront of technology development. Maria asserts that efforts to automate or standardize complex human actions and decisions are misguided and could lead to dehumanizing outcomes. They both advocate for using AI as an aid to enhance human capabilities rather than a substitute.In closing, Maria encourages a balanced approach to AI adoption, urging stakeholders to prioritize human well-being, ethical standards, and societal benefit above mere technological advancement. The conversation ends with Maria pointing people to her book for more in-depth analysis and thoughts on the future interaction between humans and technology.Buy Maria's book here: amzn.to/4avF6kqlinkedin.com/in/mariasantacaterinaPOD: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/CAN-MACHINES-REPLACE-US--AI-vs-Humanity---Maria-Santacaterina-e2jaa4pTOC00:00:00 - Intro to Book00:03:23 - What Life Is00:10:10 - Agency00:18:04 - Tech and Society00:21:51 - System 1 and 200:22:59 - We Are Being Pigeonholed00:30:22 - Agency vs Autonomy00:36:37 - Explanations00:40:24 - AI Reductionism00:49:50 - How Are Humans Intelligent01:00:22 - Semantics01:01:53 - Emotive AI and Pavlovian Dogs01:04:05 - Technology, Social Media and Organisation01:18:34 - Systems Are Not That Automated01:19:33 - Hiring01:22:34 - Subjectivity in Orgs01:32:28 - The AGI Delusion01:45:37 - GPT-laziness Syndrome01:54:58 - Diversity Preservation01:58:24 - Ethics02:11:43 - Moral Realism02:16:17 - Utopia02:18:02 - Reciprocity02:20:52 - Tyranny of CategorisationInterviewer: Dr. Tim Scarfe
What’s the Magic Word? A Control Theory of LLM Prompting #artificialintelligence Machine Learning Street Talk 2024-05-05 | "What's the Magic Word? A Control Theory of LLM Prompting"Aman Bhargava, Cameron Witkowski, Manav Shah, Matt Thomsonarxiv.org/abs/2310.04444Video out on Patreon today!
Dr. THOMAS PARR - Active Inference Machine Learning Street Talk 2024-05-01 | Thomas Parr and his collaborators wrote a book titled "Active Inference: The Free Energy Principle in Mind, Brain and Behavior" which introduces Active Inference from both a high-level conceptual perspective and a low-level mechanistic, mathematical perspective.Active inference, developed by the legendary neuroscientist Prof. Karl Friston - is a unifying mathematical framework which frames living systems as agents which minimize surprise and free energy in order to resist entropy and persist over time. It unifies various perspectives from physics, biology, statistics, and psychology - and allows us to explore deep questions about agency, biology, causality, modelling, and consciousness.Buy Active Inference: The Free Energy Principle in Mind, Brain, and BehaviorThomas Parr, Giovanni Pezzulo, Karl Friston amzn.to/4dj0iMjPlease support us on Patreon to get access to the private Discord server, bi-weekly calls, early access and ad-free listening. patreon.com/mlstPod: podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/Dr--Thomas-Parr---Active-Inference-Book-e2j4sdnTOC:00:00:00 Intro00:05:10 When Thomas met Friston00:06:13 ChatGPT comparison00:08:40 Do NNs learn a world model?00:11:04 Book intro00:13:22 High road low road of Active Inference00:17:16 Resisting entropic forces00:20:51 Agency vs free will00:26:01 Are agents real? non-physical agents00:35:54 Mind is flat / predictive brain00:44:23 Volition00:50:26 Externalism00:51:57 Bridge with Enactivism00:53:27 Bayesian Surprise01:01:47 Variational inference01:05:47 Why Bayesian?01:12:04 Causality01:17:35 Hand crafted models01:26:45 Chapter 10 - bringing it together01:28:58 Consciousness01:33:10 Humans are incoherent01:35:25 Experience writing a bookInterviewer: Dr. Tim Scarfe
Connor Leahy - e/acc, AGI and the future. Machine Learning Street Talk 2024-04-21 | Connor is the CEO of Conjecture and one of the most famous names in the AI alignment movement. This is the "behind the scenes footage" and bonus Patreon interviews from the day of the Beff Jezos debate, including an interview with Daniel Clothiaux. It's a great insight into Connor's philosophy. Support MLST:Please support us on Patreon. We are entirely funded from Patreon donations right now. Patreon supports get private discord access, biweekly calls, very early-access + exclusive content and lots more. patreon.com/mlstDonate: paypal.com/donate/?hosted_button_id=K2TYRVPBGXVNAIf you would like to sponsor us, so we can tell your story - reach out on mlstreettalk at gmailTopics:Externalized cognition and the role of society and culture in human intelligenceThe potential for AI systems to develop agency and autonomyThe future of AGI as a complex mixture of various componentsThe concept of agency and its relationship to powerThe importance of coherence in AI systemsThe balance between coherence and variance in exploring potential upsidesThe role of dynamic, competent, and incorruptible institutions in handling risks and developing technologyConcerns about AI widening the gap between the haves and have-notsThe concept of equal access to opportunity and maintaining dynamism in the systemLeahy's perspective on life as a process that "rides entropy"The importance of distinguishing between epistemological, decision-theoretic, and aesthetic aspects of morality (inc ref to Hume's Guillotine)The concept of continuous agency and the idea that the first AGI will be a messy admixture of various componentsThe potential for AI systems to become more physically embedded in the futureThe challenges of aligning AI systems and the societal impacts of AI technologies like ChatGPT and BingThe importance of humility in the face of complexity when considering the future of AI and its societal implicationsTOC:00:00:00 Intro00:00:56 Connor's Philosophy00:03:53 Office Skit00:05:08 Connor on e/acc and Beff00:07:28 Intro to Daniel's Philosophy00:08:35 Connor on Entropy, Life, and Morality00:19:10 Connor on London00:20:21 Connor Office Interview00:20:46 Friston Patreon Preview00:21:48 Why Are We So Dumb?00:23:52 The Voice of the People, the Voice of God / Populism00:26:35 Mimetics00:30:03 Governance00:33:19 Agency00:40:25 Daniel Interview - Externalised Cognition, Bing GPT, AGI00:56:29 Beff + Connor Bonus Patreons InterviewDisclaimer: this video is not an endorsement of e/acc or AGI agential existential risk from us - the hosts of MLST consider both of these views to be quite extreme. We seek diverse views on the channel.
Prof. Chris Bishops NEW Deep Learning Textbook! Machine Learning Street Talk 2024-04-10 | Professor Chris Bishop is a Technical Fellow and Director at Microsoft Research AI4Science, in Cambridge. He is also Honorary Professor of Computer Science at the University of Edinburgh, and a Fellow of Darwin College, Cambridge. In 2004, he was elected Fellow of the Royal Academy of Engineering, in 2007 he was elected Fellow of the Royal Society of Edinburgh, and in 2017 he was elected Fellow of the Royal Society. Chris was a founding member of the UK AI Council, and in 2019 he was appointed to the Prime Minister’s Council for Science and Technology.At Microsoft Research, Chris oversees a global portfolio of industrial research and development, with a strong focus on machine learning and the natural sciences.Chris obtained a BA in Physics from Oxford, and a PhD in Theoretical Physics from the University of Edinburgh, with a thesis on quantum field theory. Chris's contributions to the field of machine learning have been truly remarkable. He has authored (what is arguably) the original textbook in the field - 'Pattern Recognition and Machine Learning' (PRML) which has served as an essential reference for countless students and researchers around the world, and that was his second textbook after his highly acclaimed first textbook Neural Networks for Pattern Recognition. Recently, Chris has co-authored a new book with his son, Hugh, titled 'Deep Learning: Foundations and Concepts.' This book aims to provide a comprehensive understanding of the key ideas and techniques underpinning the rapidly evolving field of deep learning. It covers both the foundational concepts and the latest advances, making it an invaluable resource for newcomers and experienced practitioners alike.Buy Chris' textbook here:amzn.to/3vvLcChMore about Prof. Chris Bishop:en.wikipedia.org/wiki/Christopher_Bishopmicrosoft.com/en-us/research/people/cmbishopSupport MLST:Please support us on Patreon. We are entirely funded from Patreon donations right now. Patreon supports get private discord access, biweekly calls, early-access + exclusive content and lots more. patreon.com/mlstDonate: paypal.com/donate/?hosted_button_id=K2TYRVPBGXVNAIf you would like to sponsor us, so we can tell your story - reach out on mlstreettalk at gmailTOC:00:00:00 - Intro to Chris00:06:54 - Changing Landscape of AI00:08:16 - Symbolism00:09:32 - PRML00:11:02 - Bayesian Approach00:14:49 - Are NNs One Model or Many, Special vs General00:20:04 - Can Language Models Be Creative00:22:35 - Sparks of AGI00:25:52 - Creativity Gap in LLMs00:35:40 - New Deep Learning Book00:39:01 - Favourite Chapters00:44:11 - Probability Theory00:45:42 - AI4Science00:48:31 - Inductive Priors00:58:52 - Drug Discovery01:05:19 - Foundational Bias Models01:07:46 - How Fundamental Is Our Physics Knowledge?01:12:05 - Transformers01:12:59 - Why Does Deep Learning Work?01:16:59 - Inscrutability of NNs01:18:01 - Example of Simulator01:21:09 - Control
AI AGENCY ISNT HERE YET... (Dr. Philip Ball) Machine Learning Street Talk 2024-04-07 | Dr. Philip Ball is a freelance science writer. He just wrote a book called "How Life Works", discussing the how the science of Biology has advanced in the last 20 years. We focus on the concept of Agency in particular. He trained as a chemist at the University of Oxford, and as a physicist at the University of Bristol. He worked previously at Nature for over 20 years, first as an editor for physical sciences and then as a consultant editor. His writings on science for the popular press have covered topical issues ranging from cosmology to the future of molecular biology.Philip is the author of many popular books on science, including H2O: A Biography of Water, Bright Earth: The Invention of Colour, The Music Instinct and Curiosity: How Science Became Interested in Everything. His book Critical Mass won the 2005 Aventis Prize for Science Books, while Serving the Reich was shortlisted for the Royal Society Winton Science Book Prize in 2014.This is one of Tim's personal favourite MLST shows, so we have designated it a special edition. Enjoy!Buy Philip's book "How Life Works" here: amzn.to/3vSmNqpTOC:00:00:00 Outside interview00:09:17 Nativism / capacities00:11:50 Generative AI00:17:59 Inscrutability and agency00:22:06 Agency on creativity00:26:38 Could we make an agential GPT-4?00:31:06 If it agential if you tell the agents what to do00:35:40 What is agency?00:44:29 Are agents real00:48:32 Causality and agency00:54:40 Ghost in the machine / intelligibility01:00:11 Multi scale / organisation view01:04:05 Collective intelligence01:09:00 Canalisation01:13:36 Intelligence is specialised01:16:29 No free lunch01:18:19 Super intelligence01:22:05 Mind is flat / confabulated goals01:25:07 Is planning/goals explicit?01:30:17 Sentience in LLMs01:34:56 Are LLMs simulators?01:40:04 Could LLMs feel?01:49:33 Digital physics view01:51:46 Agential vs nonagential AI01:54:02 Bostrom thinks goals and intelligence are separate02:05:41 Simulation sharingAI Transcript: docs.google.com/document/d/164FMAswgS9jAqxAaB5fkUtrJKMdq8rW7b94XgibvQ7Q/edit?usp=sharingSupport MLST:Please support us on Patreon. We are entirely funded from Patreon donations right now. Patreon supports get private discord access, biweekly calls, early-access + exclusive content and lots more. patreon.com/mlstDonate: paypal.com/donate/?hosted_button_id=K2TYRVPBGXVNAIf you would like to sponsor us, so we can tell your story - reach out on mlstreettalk at gmail
