Machine Learning Street Talk (MLST)

Please check out Numerai - our sponsor @

https://numerai.com/mlst

Numerai is a groundbreaking platform which is taking the data science world by storm. Tim has been using Numerai to build state-of-the-art models which predict the stock market, all while being a part of an inspiring community of data scientists from around the globe. They host the Numerai Data Science Tournament, where data scientists like us use their financial dataset to predict future stock market performance.

Support us! https://www.patreon.com/mlst

MLST Discord: https://discord.gg/aNPkGUQtc5

Twitter: https://twitter.com/MLStreetTalk

Welcome to an exciting episode featuring an outstanding guest, Robert Miles! Renowned for his extraordinary contributions to understanding AI and its potential impacts on our lives, Robert is an artificial intelligence advocate, researcher, and YouTube sensation. He combines engaging discussions with entertaining content, captivating millions of viewers from around the world.

With a strong computer science background, Robert has been actively involved in AI safety projects, focusing on raising awareness about potential risks and benefits of advanced AI systems. His YouTube channel is celebrated for making AI safety discussions accessible to a diverse audience through breaking down complex topics into easy-to-understand nuggets of knowledge, and you might also recognise him from his appearances on Computerphile.

In this episode, join us as we dive deep into Robert's journey in the world of AI, exploring his insights on AI alignment, superintelligence, and the role of AI shaping our society and future. We'll discuss topics such as the limits of AI capabilities and physics, AI progress and timelines, human-machine hybrid intelligence, AI in conflict and cooperation with humans, and the convergence of AI communities.

Robert Miles:

@RobertMilesAI

https://twitter.com/robertskmiles

https://aisafety.info/

YT version: https://www.youtube.com/watch?v=kMLKbhY0ji0

Panel:

Dr. Tim Scarfe

Dr. Keith Duggar

Joint CTOs - https://xrai.glass/

Refs:

Are Emergent Abilities of Large Language Models a Mirage? (Rylan Schaeffer)

https://arxiv.org/abs/2304.15004

TOC:

Intro [00:00:00]

Numerai Sponsor Messsage [00:02:17]

AI Alignment [00:04:27]

Limits of AI Capabilities and Physics [00:18:00]

AI Progress and Timelines [00:23:52]

AI Arms Race and Innovation [00:31:11]

Human-Machine Hybrid Intelligence [00:38:30]

Understanding and Defining Intelligence [00:42:48]

AI in Conflict and Cooperation with Humans [00:50:13]

Interpretability and Mind Reading in AI [01:03:46]

Mechanistic Interpretability and Deconfusion Research [01:05:53]

Understanding the core concepts of AI [01:07:40]

Moon landing analogy and AI alignment [01:09:42]

Cognitive horizon and limits of human intelligence [01:11:42]

Funding and focus on AI alignment [01:16:18]

Regulating AI technology and potential risks [01:19:17]

Aligning AI with human values and its dynamic nature [01:27:04]

Cooperation and Allyship [01:29:33]

Orthogonality Thesis and Goal Preservation [01:33:15]

Anthropomorphic Language and Intelligent Agents [01:35:31]

Maintaining Variety and Open-ended Existence [01:36:27]

Emergent Abilities of Large Language Models [01:39:22]

Convergence vs Emergence [01:44:04]

Criticism of X-risk and Alignment Communities [01:49:40]

Fusion of AI communities and addressing biases [01:52:51]

AI systems integration into society and understanding them [01:53:29]

Changing opinions on AI topics and learning from past videos [01:54:23]

Utility functions and von Neumann-Morgenstern theorems [01:54:47]

AI Safety FAQ project [01:58:06]

Building a conversation agent using AI safety dataset [02:00:36]

Generative Deep Learning, 2nd Edition [David Foster]

https://www.oreilly.com/library/view/generative-deep-learning/9781098134174/

Support us! https://www.patreon.com/mlst

MLST Discord: https://discord.gg/aNPkGUQtc5

Twitter: https://twitter.com/MLStreetTalk

In this conversation, Tim Scarfe and David Foster, the author of 'Generative Deep Learning,' dive deep into the world of generative AI, discussing topics ranging from model families and auto regressive models to the democratization of AI technology and its potential impact on various industries. They explore the connection between language and true intelligence, as well as the limitations of GPT and other large language models. The discussion also covers the importance of task-independent world models, the concept of active inference, and the potential of combining these ideas with transformer and GPT-style models.

Ethics and regulation in AI development are also discussed, including the need for transparency in data used to train AI models and the responsibility of developers to ensure their creations are not destructive. The conversation touches on the challenges posed by AI-generated content on copyright laws and the diminishing role of effort and skill in copyright due to generative models.

The impact of AI on education and creativity is another key area of discussion, with Tim and David exploring the potential benefits and drawbacks of using AI in the classroom, the need for a balance between traditional learning methods and AI-assisted learning, and the importance of teaching students to use AI tools critically and responsibly.

Generative AI in music is also explored, with David and Tim discussing the potential for AI-generated music to change the way we create and consume art, as well as the challenges in training AI models to generate music that captures human emotions and experiences.

Throughout the conversation, Tim and David touch on the potential risks and consequences of AI becoming too powerful, the importance of maintaining control over the technology, and the possibility of government intervention and regulation. The discussion concludes with a thought experiment about AI predicting human actions and creating transient capabilities that could lead to doom.

TOC:

Introducing Generative Deep Learning [00:00:00]

Model Families in Generative Modeling [00:02:25]

Auto Regressive Models and Recurrence [00:06:26]

Language and True Intelligence [00:15:07]

Language, Reality, and World Models [00:19:10]

AI, Human Experience, and Understanding [00:23:09]

GPTs Limitations and World Modeling [00:27:52]

Task-Independent Modeling and Cybernetic Loop [00:33:55]

Collective Intelligence and Emergence [00:36:01]

Active Inference vs. Reinforcement Learning [00:38:02]

Combining Active Inference with Transformers [00:41:55]

Decentralized AI and Collective Intelligence [00:47:46]

Regulation and Ethics in AI Development [00:53:59]

AI-Generated Content and Copyright Laws [00:57:06]

Effort, Skill, and AI Models in Copyright [00:57:59]

AI Alignment and Scale of AI Models [00:59:51]

Democratization of AI: GPT-3 and GPT-4 [01:03:20]

Context Window Size and Vector Databases [01:10:31]

Attention Mechanisms and Hierarchies [01:15:04]

Benefits and Limitations of Language Models [01:16:04]

AI in Education: Risks and Benefits [01:19:41]

AI Tools and Critical Thinking in the Classroom [01:29:26]

Impact of Language Models on Assessment and Creativity [01:35:09]

Generative AI in Music and Creative Arts [01:47:55]

Challenges and Opportunities in Generative Music [01:52:11]

AI-Generated Music and Human Emotions [01:54:31]

Language Modeling vs. Music Modeling [02:01:58]

Democratization of AI and Industry Impact [02:07:38]

Recursive Self-Improving Superintelligence [02:12:48]

AI Technologies: Positive and Negative Impacts [02:14:44]

Runaway AGI and Control Over AI [02:20:35]

AI Dangers, Cybercrime, and Ethics [02:23:42]

https://www.perplexity.ai/

https://www.perplexity.ai/iphone

https://www.perplexity.ai/android Interview with Aravind Srinivas, CEO and Co-Founder of Perplexity AI – Revolutionizing Learning with Conversational Search Engines Dr. Tim Scarfe talks with Dr. Aravind Srinivas, CEO and Co-Founder of Perplexity AI, about his journey from studying AI and reinforcement learning at UC Berkeley to launching Perplexity – a startup that aims to revolutionize learning through the power of conversational search engines. By combining the strengths of large language models like GPT-* with search engines, Perplexity provides users with direct answers to their questions in a decluttered user interface, making the learning process not only more efficient but also enjoyable. Aravind shares his insights on how advertising can be made more relevant and less intrusive with the help of large language models, emphasizing the importance of transparency in relevance ranking to improve user experience. He also discusses the challenge of balancing the interests of users and advertisers for long-term success. The interview delves into the challenges of maintaining truthfulness and balancing opinions and facts in a world where algorithmic truth is difficult to achieve. Aravind believes that opinionated models can be useful as long as they don't spread misinformation and are transparent about being opinions. He also emphasizes the importance of allowing users to correct or update information, making the platform more adaptable and dynamic. Lastly, Aravind shares his thoughts on embracing a digital society with large language models, stressing the need for frequent and iterative deployments of these models to reduce fear of AI and misinformation. He envisions a future where using AI tools effectively requires clear thinking and first-principle reasoning, ultimately benefiting society as a whole. Education and transparency are crucial to counter potential misuse of AI for political or malicious purposes.

YT version: https://youtu.be/_vMOWw3uYvk Aravind Srinivas: https://www.linkedin.com/in/aravind-srinivas-16051987/

https://scholar.google.com/citations?user=GhrKC1gAAAAJ&hl=en

https://twitter.com/aravsrinivas?lang=en Interviewer: Dr. Tim Scarfe (CTO XRAI Glass) Patreon: https://www.patreon.com/mlst Discord: https://discord.gg/ESrGqhf5CB TOC: Introduction and Background of Perplexity AI [00:00:00]

The Importance of a Decluttered UI and User Experience [00:04:19]

Advertising in Search Engines and Potential Improvements [00:09:02]

Challenges and Opportunities in this new Search Modality [00:18:17]

Benefits of Perplexity and Personalized Learning [00:21:27]

Objective Truth and Personalized Wikipedia [00:26:34]

Opinions and Truth in Answer Engines [00:30:53]

Embracing the Digital Society with Language Models [00:37:30]

Impact on Jobs and Future of Learning [00:40:13]

Educating users on when perplexity works and doesn't work [00:43:13]

Improving user experience and the possibilities of voice-to-voice interaction [00:45:04]

The future of language models and auto-regressive models [00:49:51]

Performance of GPT-4 and potential improvements [00:52:31]

Building the ultimate research and knowledge assistant [00:55:33]

Revolutionizing note-taking and personal knowledge stores [00:58:16] References: Evaluating Verifiability in Generative Search Engines (Nelson F. Liu et al, Stanford University) https://arxiv.org/pdf/2304.09848.pdf Note: this was a sponsored interview.

Support us! https://www.patreon.com/mlst MLST Discord: https://discord.gg/aNPkGUQtc5 In this podcast with the legendary Connor Leahy (CEO Conjecture) recorded in Dec 2022, we discuss various topics related to artificial intelligence (AI), including AI alignment, the success of ChatGPT, the potential threats of artificial general intelligence (AGI), and the challenges of balancing research and product development at his company, Conjecture. He emphasizes the importance of empathy, dehumanizing our thinking to avoid anthropomorphic biases, and the value of real-world experiences in learning and personal growth. The conversation also covers the Orthogonality Thesis, AI preferences, the mystery of mode collapse, and the paradox of AI alignment. Connor Leahy expresses concern about the rapid development of AI and the potential dangers it poses, especially as AI systems become more powerful and integrated into society. He argues that we need a better understanding of AI systems to ensure their safe and beneficial development. The discussion also touches on the concept of "futuristic whack-a-mole," where futurists predict potential AGI threats, and others try to come up with solutions for those specific scenarios. However, the problem lies in the fact that there could be many more scenarios that neither party can think of, especially when dealing with a system that's smarter than humans. https://www.linkedin.com/in/connor-j-leahy/https://twitter.com/NPCollapse Interviewer: Dr. Tim Scarfe (Innovation CTO @ XRAI Glass https://xrai.glass/) TOC: The success of ChatGPT and its impact on the AI field [00:00:00] Subjective experience [00:15:12] AI Architectural discussion including RLHF [00:18:04] The paradox of AI alignment and the future of AI in society [00:31:44] The impact of AI on society and politics [00:36:11] Future shock levels and the challenges of predicting the future [00:45:58] Long termism and existential risk [00:48:23] Consequentialism vs. deontology in rationalism [00:53:39] The Rationalist Community and its Challenges [01:07:37] AI Alignment and Conjecture [01:14:15] Orthogonality Thesis and AI Preferences [01:17:01] Challenges in AI Alignment [01:20:28] Mechanistic Interpretability in Neural Networks [01:24:54] Building Cleaner Neural Networks [01:31:36] Cognitive horizons / The problem with rapid AI development [01:34:52] Founding Conjecture and raising funds [01:39:36] Inefficiencies in the market and seizing opportunities [01:45:38] Charisma, authenticity, and leadership in startups [01:52:13] Autistic culture and empathy [01:55:26] Learning from real-world experiences [02:01:57] Technical empathy and transhumanism [02:07:18] Moral status and the limits of empathy [02:15:33] Anthropomorphic Thinking and Consequentialism [02:17:42] Conjecture: Balancing Research and Product Development [02:20:37] Epistemology Team at Conjecture [02:31:07] Interpretability and Deception in AGI [02:36:23] Futuristic whack-a-mole and predicting AGI threats [02:38:27] Refs: 1. OpenAI's ChatGPT: https://chat.openai.com/ 2. The Mystery of Mode Collapse (Article): https://www.lesswrong.com/posts/t9svvNPNmFf5Qa3TA/mysteries-of-mode-collapse 3. The Rationalist Guide to the Galaxy https://www.amazon.co.uk/Does-Not-Hate-You-Superintelligence/dp/1474608795 5. Alfred Korzybski: https://en.wikipedia.org/wiki/Alfred_Korzybski 6. Instrumental Convergence: https://en.wikipedia.org/wiki/Instrumental_convergence 7. Orthogonality Thesis: https://en.wikipedia.org/wiki/Orthogonality_thesis 8. Brian Tomasik's Essays on Reducing Suffering: https://reducing-suffering.org/ 9. Epistemological Framing for AI Alignment Research: https://www.lesswrong.com/posts/Y4YHTBziAscS5WPN7/epistemological-framing-for-ai-alignment-research 10. How to Defeat Mind readers: https://www.alignmentforum.org/posts/EhAbh2pQoAXkm9yor/circumventing-interpretability-how-to-defeat-mind-readers 11. Society of mind: https://www.amazon.co.uk/Society-Mind-Marvin-Minsky/dp/0671607405

Support us! https://www.patreon.com/mlst MLST Discord: https://discord.gg/aNPkGUQtc5

Send us a voice message which you want us to publish: https://podcasters.spotify.com/pod/show/machinelearningstreettalk/message In a recent open letter, over 1500 individuals called for a six-month pause on the development of advanced AI systems, expressing concerns over the potential risks AI poses to society and humanity. However, there are issues with this approach, including global competition, unstoppable progress, potential benefits, and the need to manage risks instead of avoiding them. Decision theorist Eliezer Yudkowsky took it a step further in a Time magazine article, calling for an indefinite and worldwide moratorium on Artificial General Intelligence (AGI) development, warning of potential catastrophe if AGI exceeds human intelligence. Yudkowsky urged for an immediate halt to all large AI training runs and the shutdown of major GPU clusters, calling for international cooperation to enforce these measures. However, several counterarguments question the validity of Yudkowsky's concerns:

1. Hard limits on AGI 2. Dismissing AI extinction risk 3. Collective action problem 4. Misplaced focus on AI threats While the potential risks of AGI cannot be ignored, it is essential to consider various arguments and potential solutions before making drastic decisions. As AI continues to advance, it is crucial for researchers, policymakers, and society as a whole to engage in open and honest discussions about the potential consequences and the best path forward. With a balanced approach to AGI development, we may be able to harness its power for the betterment of humanity while mitigating its risks. Eliezer Yudkowsky: https://en.wikipedia.org/wiki/Eliezer_Yudkowsky Connor Leahy: https://twitter.com/NPCollapse (we will release that interview soon) Gary Marcus: http://garymarcus.com/index.html Tim Scarfe is the innovation CTO of XRAI Glass: https://xrai.glass/ Gary clip filmed at AIUK https://ai-uk.turing.ac.uk/programme/ and our appreciation to them for giving us a press pass. Check out their conference next year! WIRED clip from Gary came from here: https://www.youtube.com/watch?v=Puo3VkPkNZ4 Refs:

Statement from the listed authors of Stochastic Parrots on the “AI pause” letterTimnit Gebru, Emily M. Bender, Angelina McMillan-Major, Margaret Mitchell

https://www.dair-institute.org/blog/letter-statement-March2023 Eliezer Yudkowsky on Lex: https://www.youtube.com/watch?v=AaTRHFaaPG8 Pause Giant AI Experiments: An Open Letter https://futureoflife.org/open-letter/pause-giant-ai-experiments/ Pausing AI Developments Isn't Enough. We Need to Shut it All Down (Eliezer Yudkowsky) https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough/

HUGE ANNOUNCEMENT, CHATGPT+WOLFRAM! You saw it HERE first! YT version: https://youtu.be/z5WZhCBRDpU Support us! https://www.patreon.com/mlst

MLST Discord: https://discord.gg/aNPkGUQtc5 Stephen's announcement post: https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-superpowers/ OpenAI's announcement post: https://openai.com/blog/chatgpt-plugins In an era of technology and innovation, few individuals have left as indelible a mark on the fabric of modern science as our esteemed guest, Dr. Steven Wolfram. Dr. Wolfram is a renowned polymath who has made significant contributions to the fields of physics, computer science, and mathematics. A prodigious young man too, Wolfram earned a Ph.D. in theoretical physics from the California Institute of Technology by the age of 20. He became the youngest recipient of the prestigious MacArthur Fellowship at the age of 21. Wolfram's groundbreaking computational tool, Mathematica, was launched in 1988 and has become a cornerstone for researchers and innovators worldwide. In 2002, he published "A New Kind of Science," a paradigm-shifting work that explores the foundations of science through the lens of computational systems. In 2009, Wolfram created Wolfram Alpha, a computational knowledge engine utilized by millions of users worldwide. His current focus is on the Wolfram Language, a powerful programming language designed to democratize access to cutting-edge technology. Wolfram's numerous accolades include honorary doctorates and fellowships from prestigious institutions. As an influential thinker, Dr. Wolfram has dedicated his life to unraveling the mysteries of the universe and making computation accessible to all. First of all... we have an announcement to make, you heard it FIRST here on MLST! .... Intro [00:00:00] Big announcement! Wolfram + ChatGPT! [00:02:57] What does it mean to understand? [00:05:33] Feeding information back into the model [00:13:48] Semantics and cognitive categories [00:20:09] Navigating the ruliad [00:23:50] Computational irreducibility [00:31:39] Conceivability and interestingness [00:38:43] Human intelligible sciences [00:43:43]

YT version: https://youtu.be/P1j3VoKBxbc (references in pinned comment) Support us! https://www.patreon.com/mlst MLST Discord: https://discord.gg/aNPkGUQtc5 Dan McQuillan, a visionary in digital culture and social innovation, emphasizes the importance of understanding technology's complex relationship with society. As an academic at Goldsmiths, University of London, he fosters interdisciplinary collaboration and champions data-driven equity and ethical technology. Dan's career includes roles at Amnesty International and Social Innovation Camp, showcasing technology's potential to empower and bring about positive change. In this conversation, we discuss the challenges and opportunities at the intersection of technology and society, exploring the profound impact of our digital world. Interviewer: Dr. Tim Scarfe

[00:00:00] Dan's background and journey to academia

[00:03:30] Dan's background and journey to academia

[00:04:10] Writing the book "Resisting AI"

[00:08:30] Necropolitics and its relation to AI

[00:10:06] AI as a new form of colonization

[00:12:57] LLMs as a new form of neo-techno-imperialism

[00:15:47] Technology for good and AGI's skewed worldview

[00:17:49] Transhumanism, eugenics, and intelligence

[00:20:45] Valuing differences (disability) and challenging societal norms

[00:26:08] Re-ontologizing and the philosophy of information

[00:28:19] New materialism and the impact of technology on society

[00:30:32] Intelligence, meaning, and materiality

[00:31:43] The constraints of physical laws and the importance of science

[00:32:44] Exploring possibilities to reduce suffering and increase well-being

[00:33:29] The division between meaning and material in our experiences

[00:35:36] Machine learning, data science, and neoplatonic approach to understanding reality

[00:37:56] Different understandings of cognition, thought, and consciousness

[00:39:15] Enactivism and its variants in cognitive science

[00:40:58] Jordan Peterson

[00:44:47] Relationism, relativism, and finding the correct relational framework

[00:47:42] Recognizing privilege and its impact on social interactions

[00:49:10] Intersectionality / Feminist thinking and the concept of care in social structures

[00:51:46] Intersectionality and its role in understanding social inequalities

[00:54:26] The entanglement of history, technology, and politics

[00:57:39] ChatGPT article - we come to bury ChatGPT

[00:59:41] Statistical pattern learning and convincing patterns in AI

[01:01:27] Anthropomorphization and understanding in AI

[01:03:26] AI in education and critical thinking

[01:06:09] European Union policies and trustable AI

[01:07:52] AI reliability and the halo effect

[01:09:26] AI as a tool enmeshed in society

[01:13:49] Luddites

[01:15:16] AI is a scam

[01:15:31] AI and Social Relations

[01:16:49] Invisible Labor in AI and Machine Learning

[01:21:09] Exploititative AI / alignment

[01:23:50] Science fiction AI / moral frameworks

[01:27:22] Discussing Stochastic Parrots and Nihilism

[01:30:36] Human Intelligence vs. Language Models

[01:32:22] Image Recognition and Emulation vs. Experience

[01:34:32] Thought Experiments and Philosophy in AI Ethics (mimicry)

[01:41:23] Abstraction, reduction, and grounding in reality

[01:43:13] Process philosophy and the possibility of change

[01:49:55] Mental health, AI, and epistemic injustice

[01:50:30] Hermeneutic injustice and gendered techniques

[01:53:57] AI and politics

[01:59:24] Epistemic injustice and testimonial injustice

[02:11:46] Fascism and AI discussion

[02:13:24] Violence in various systems

[02:16:52] Recognizing systemic violence

[02:22:35] Fascism in Today's Society

[02:33:33] Pace and Scale of Technological Change

[02:37:38] Alternative approaches to AI and society

[02:44:09] Self-Organization at Successive Scales / cybernetics

Support us! https://www.patreon.com/mlst  

MLST Discord: https://discord.gg/aNPkGUQtc5

We are honoured to welcome Dr. Joel Lehman, an eminent machine learning research scientist, whose work in AI safety, reinforcement learning, creative open-ended search algorithms, and indeed the philosophy of open-endedness and abandoning objectives has paved the way for innovative ideas that challenge our preconceptions and inspire new visions for the future.

Dr. Lehman's thought-provoking book, "Why Greatness Cannot Be Planned" penned with with our MLST favourite Professor Kenneth Stanley has left an indelible mark on the field and profoundly impacted the way we view innovation and the serendipitous nature of discovery. Those of you who haven't watched our special edition show on that, should do so at your earliest convenience! Building upon this foundation, Dr. Lehman has ventured into the domain of AI systems that embody principles of love, care, responsibility, respect, and knowledge, drawing from the works of Maslow, Erich Fromm, and positive psychology.

YT version: https://youtu.be/23-TXgJEv-Q

http://joellehman.com/

https://twitter.com/joelbot3000

Interviewer: Dr. Tim Scarfe

TOC:

Intro [00:00:00]

Model [00:04:26]

Intro and Paper Intro [00:08:52]

Subjectivity [00:16:07]

Reflections on Greatness Book [00:19:30]

Representing Subjectivity [00:29:24]

Nagal's Bat [00:31:49]

Abstraction [00:38:58]

Love as Action Rather Than Feeling [00:42:58]

Reontologisation [00:57:38]

Self Help [01:04:15]

Meditation [01:09:02]

The Human Reward Function / Effective... [01:16:52]

Machine Hate [01:28:32]

Societal Harms [01:31:41]

Lenses We Use Obscuring Reality [01:56:36]

Meta Optimisation and Evolution [02:03:14]

Conclusion [02:07:06]

References:

What Is It Like to Be a Bat? (Thomas Nagel)

https://warwick.ac.uk/fac/cross_fac/iatl/study/ugmodules/humananimalstudies/lectures/32/nagel_bat.pdf

Why Greatness Cannot Be Planned: The Myth of the Objective (Kenneth O. Stanley and Joel Lehman)

https://link.springer.com/book/10.1007/978-3-319-15524-1 

Machine Love (Joel Lehman)

https://arxiv.org/abs/2302.09248 

How effective altruists ignored risk (Carla Cremer)

https://www.vox.com/future-perfect/23569519/effective-altrusim-sam-bankman-fried-will-macaskill-ea-risk-decentralization-philanthropy

Philosophy tube - The Rich Have Their Own Ethics: Effective Altruism

https://www.youtube.com/watch?v=Lm0vHQYKI-Y

Abandoning Objectives: Evolution through the Search for Novelty Alone (Joel Lehman and Kenneth O. Stanley)

https://www.cs.swarthmore.edu/~meeden/DevelopmentalRobotics/lehman_ecj11.pdf

Support us! https://www.patreon.com/mlst

MLST Discord: https://discord.gg/aNPkGUQtc5

Dr. Raphaël Millière is the 2020 Robert A. Burt Presidential Scholar in Society and Neuroscience in the Center for Science and Society, and a Lecturer in the Philosophy Department at Columbia University. His research draws from his expertise in philosophy and cognitive science to explore the implications of recent progress in deep learning for models of human cognition, as well as various issues in ethics and aesthetics. He is also investigating what underlies the capacity to represent oneself as oneself at a fundamental level, in humans and non-human animals; as well as the role that self-representation plays in perception, action, and memory. In a world where technology is rapidly advancing, Dr. Millière is striving to gain a better understanding of how artificial neural networks work, and to establish fair and meaningful comparisons between humans and machines in various domains in order to shed light on the implications of artificial intelligence for our lives.

https://www.raphaelmilliere.com/

https://twitter.com/raphaelmilliere

Here is a version with hesitation sounds like "um" removed if you prefer (I didn't notice them personally): https://share.descript.com/view/aGelyTl2xpN

YT: https://www.youtube.com/watch?v=fhn6ZtD6XeE

TOC:

Intro to Raphael [00:00:00]

Intro: Moving Beyond Mimicry in Artificial Intelligence (Raphael Millière) [00:01:18]

Show Kick off [00:07:10]

LLMs [00:08:37]

Semantic Competence/Understanding [00:18:28]

Forming Analogies/JPG Compression Article [00:30:17]

Compositional Generalisation [00:37:28]

Systematicity [00:47:08]

Language of Thought [00:51:28]

Bigbench (Conceptual Combinations) [00:57:37]

Symbol Grounding [01:11:13]

World Models [01:26:43]

Theory of Mind [01:30:57]

Refs (this is truncated, full list on YT video description):

Moving Beyond Mimicry in Artificial Intelligence (Raphael Millière)

https://nautil.us/moving-beyond-mimicry-in-artificial-intelligence-238504/

On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜 (Bender et al)

https://dl.acm.org/doi/10.1145/3442188.3445922

ChatGPT Is a Blurry JPEG of the Web (Ted Chiang)

https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web

The Debate Over Understanding in AI's Large Language Models (Melanie Mitchell)

https://arxiv.org/abs/2210.13966

Talking About Large Language Models (Murray Shanahan)

https://arxiv.org/abs/2212.03551

Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data (Bender)

https://aclanthology.org/2020.acl-main.463/

The symbol grounding problem (Stevan Harnad)

https://arxiv.org/html/cs/9906002

Why the Abstraction and Reasoning Corpus is interesting and important for AI (Mitchell)

https://aiguide.substack.com/p/why-the-abstraction-and-reasoning

Linguistic relativity (Sapir–Whorf hypothesis)

https://en.wikipedia.org/wiki/Linguistic_relativity

Cooperative principle (Grice's four maxims of conversation - quantity, quality, relation, and manner)

https://en.wikipedia.org/wiki/Cooperative_principle

This show is sponsored by Numerai, please visit them here with our sponsor link (we would really appreciate it) http://numer.ai/mlst 

Prof. Karl Friston recently proposed a vision of artificial intelligence that goes beyond machines and algorithms, and embraces humans and nature as part of a cyber-physical ecosystem of intelligence. This vision is based on the principle of active inference, which states that intelligent systems can learn from their observations and act on their environment to reduce uncertainty and achieve their goals. This leads to a formal account of collective intelligence that rests on shared narratives and goals. 

To realize this vision, Friston suggests developing a shared hyper-spatial modelling language and transaction protocol, as well as novel methods for measuring and optimizing collective intelligence. This could harness the power of artificial intelligence for the common good, without compromising human dignity or autonomy. It also challenges us to rethink our relationship with technology, nature, and each other, and invites us to join a global community of sense-makers who are curious about the world and eager to improve it.

YT version: https://www.youtube.com/watch?v=V_VXOdf1NMw

Support us! https://www.patreon.com/mlst 

MLST Discord: https://discord.gg/aNPkGUQtc5

TOC: 

Intro [00:00:00]

Numerai (Sponsor segment) [00:07:10]

Designing Ecosystems of Intelligence from First Principles (Friston et al) [00:09:48]

Information / Infosphere and human agency [00:18:30]

Intelligence [00:31:38]

Reductionism [00:39:36]

Universalism [00:44:46]

Emergence [00:54:23]

Markov blankets [01:02:11]

Whole part relationships / structure learning [01:22:33]

Enactivism [01:29:23]

Knowledge and Language [01:43:53]

ChatGPT [01:50:56]

Ethics (is-ought) [02:07:55]

Can people be evil? [02:35:06]

Ethics in Al, subjectiveness [02:39:05]

Final thoughts [02:57:00]

References:

Designing Ecosystems of Intelligence from First Principles (Friston et al)

https://arxiv.org/abs/2212.01354

GLOM - How to represent part-whole hierarchies in a neural network (Hinton)

https://arxiv.org/pdf/2102.12627.pdf

Seven Brief Lessons on Physics (Carlo Rovelli)

https://www.amazon.co.uk/Seven-Brief-Lessons-Physics-Rovelli/dp/0141981725

How Emotions Are Made: The Secret Life of the Brain (Lisa Feldman Barrett)

https://www.amazon.co.uk/How-Emotions-Are-Made-Secret/dp/B01N3D4OON

Am I Self-Conscious? (Or Does Self-Organization Entail Self-Consciousness?) (Karl Friston)

https://www.frontiersin.org/articles/10.3389/fpsyg.2018.00579/full

Integrated information theory (Giulio Tononi)

https://en.wikipedia.org/wiki/Integrated_information_theory

Access Numerai here: http://numer.ai/mlst

Michael Oliver is the Chief Scientist at Numerai, a hedge fund that crowdsources machine learning models from data scientists. He has a PhD in Computational Neuroscience from UC Berkeley and was a postdoctoral researcher at the Allen Institute for Brain Science before joining Numerai in 2020. He is also the host of Numerai Quant Club, a YouTube series where he discusses Numerai’s research, data and challenges.

YT version: https://youtu.be/61s8lLU7sFg

TOC:

[00:00:00] Introduction to Michael and Numerai

[00:02:03] Understanding / new Bing

[00:22:47] Quant vs Neuroscience

[00:36:43] Role of language in cognition and planning, and subjective... 

[00:45:47] Boundaries in finance modelling

[00:48:00] Numerai

[00:57:37] Aggregation systems

[01:00:52] Getting started on Numeral

[01:03:21] What models are people using

[01:04:23] Numerai Problem Setup

[01:05:49] Regimes in financial data and quant talk

[01:11:18] Esoteric approaches used on Numeral?

[01:13:59]  Curse of dimensionality

[01:16:32] Metrics

[01:19:10] Outro

References:

Growing Neural Cellular Automata (Alexander Mordvintsev)

https://distill.pub/2020/growing-ca/

A Thousand Brains: A New Theory of Intelligence (Jeff Hawkins)

https://www.amazon.fr/Thousand-Brains-New-Theory-Intelligence/dp/1541675819

Perceptual Neuroscience: The Cerebral Cortex (Vernon B. Mountcastle)

https://www.amazon.ca/Perceptual-Neuroscience-Cerebral-Vernon-Mountcastle/dp/0674661885

Numerai Quant Club with Michael Oliver

https://www.youtube.com/watch?v=eLIxarbDXuQ&list=PLz3D6SeXhT3tTu8rhZmjwDZpkKi-UPO1F

Numerai YT channel

https://www.youtube.com/@Numerai/featured

Support us! https://www.patreon.com/mlst 

MLST Discord: https://discord.gg/aNPkGUQtc5

Support us! https://www.patreon.com/mlst  

MLST Discord: https://discord.gg/aNPkGUQtc5

Christopher Summerfield, Department of Experimental Psychology, University of Oxford is a Professor of Cognitive Neuroscience at the University of Oxford and a Research Scientist at Deepmind UK. His work focusses on the neural and computational mechanisms by which humans make decisions.

Chris has just released an incredible new book on AI called "Natural General Intelligence". It's my favourite book on AI I have read so so far. 

The book explores the algorithms and architectures that are driving progress in AI research, and discusses intelligence in the language of psychology and biology, using examples and analogies to be comprehensible to a wide audience. It also tackles longstanding theoretical questions about the nature of thought and knowledge.

With Chris' permission, I read out a summarised version of Chapter 2 from his book on which was on Intelligence during the 30 minute MLST introduction.  

Buy his book here:

https://global.oup.com/academic/product/natural-general-intelligence-9780192843883?cc=gb&lang=en&

YT version: https://youtu.be/31VRbxAl3t0

Interviewer: Dr. Tim Scarfe

TOC:

[00:00:00] Walk and talk with Chris on Knowledge and Abstractions

[00:04:08] Intro to Chris and his book

[00:05:55] (Intro) Tim reads Chapter 2: Intelligence 

[00:09:28] Intro continued: Goodhart's law

[00:15:37] Intro continued: The "swiss cheese" situation  

[00:20:23] Intro continued: On Human Knowledge

[00:23:37] Intro continued: Neats and Scruffies

[00:30:22] Interview kick off 

[00:31:59] What does it mean to understand?

[00:36:18] Aligning our language models

[00:40:17] Creativity 

[00:41:40] "Meta" AI and basins of attraction 

[00:51:23] What can Neuroscience impart to AI

[00:54:43] Sutton, neats and scruffies and human alignment

[01:02:05] Reward is enough

[01:19:46] Jon Von Neumann and Intelligence

[01:23:56] Compositionality

References:

The Language Game (Morten H. Christiansen, Nick Chater

https://www.penguin.co.uk/books/441689/the-language-game-by-morten-h-christiansen-and--nick-chater/9781787633483

Theory of general factor (Spearman)

https://www.proquest.com/openview/7c2c7dd23910c89e1fc401e8bb37c3d0/1?pq-origsite=gscholar&cbl=1818401

Intelligence Reframed (Howard Gardner)

https://books.google.co.uk/books?hl=en&lr=&id=Qkw4DgAAQBAJ&oi=fnd&pg=PT6&dq=howard+gardner+multiple+intelligences&ots=ERUU0u5Usq&sig=XqiDgNUIkb3K9XBq0vNbFmXWKFs#v=onepage&q=howard%20gardner%20multiple%20intelligences&f=false

The master algorithm (Pedro Domingos)

https://www.amazon.co.uk/Master-Algorithm-Ultimate-Learning-Machine/dp/0241004543

A Thousand Brains: A New Theory of Intelligence (Jeff Hawkins)

https://www.amazon.co.uk/Thousand-Brains-New-Theory-Intelligence/dp/1541675819

The bitter lesson (Rich Sutton)

http://www.incompleteideas.net/IncIdeas/BitterLesson.html

Support us! https://www.patreon.com/mlst 

MLST Discord: https://discord.gg/aNPkGUQtc5

YT: https://youtu.be/i9VPPmQn9HQ

Edward Grefenstette is a Franco-American computer scientist who currently serves as Head of Machine Learning at Cohere and Honorary Professor at UCL. He has previously been a research scientist at Facebook AI Research and staff research scientist at DeepMind, and was also the CTO of Dark Blue Labs. Prior to his move to industry, Edward was a Fulford Junior Research Fellow at Somerville College, University of Oxford, and was lecturing at Hertford College. He obtained his BSc in Physics and Philosophy from the University of Sheffield and did graduate work in the philosophy departments at the University of St Andrews. His research draws on topics and methods from Machine Learning, Computational Linguistics and Quantum Information Theory, and has done work implementing and evaluating compositional vector-based models of natural language semantics and empirical semantic knowledge discovery.

https://www.egrefen.com/

https://cohere.ai/

TOC:

[00:00:00] Introduction

[00:02:52] Differential Semantics

[00:06:56] Concepts

[00:10:20] Ontology

[00:14:02] Pragmatics

[00:16:55] Code helps with language

[00:19:02] Montague

[00:22:13] RLHF

[00:31:54] Swiss cheese problem / retrieval augmented

[00:37:06] Intelligence / Agency

[00:43:33] Creativity

[00:46:41] Common sense

[00:53:46] Thinking vs knowing

References:

Large language models are not zero-shot communicators (Laura Ruis)

https://arxiv.org/abs/2210.14986

Some remarks on Large Language Models (Yoav Goldberg)

https://gist.github.com/yoavg/59d174608e92e845c8994ac2e234c8a9

Quantum Natural Language Processing (Bob Coecke)

https://www.cs.ox.ac.uk/people/bob.coecke/QNLP-ACT.pdf

Constitutional AI: Harmlessness from AI Feedback

https://www.anthropic.com/constitutional.pdf

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Patrick Lewis)

https://www.patricklewis.io/publication/rag/

Natural General Intelligence (Prof. Christopher Summerfield)

https://global.oup.com/academic/product/natural-general-intelligence-9780192843883

ChatGPT with Rob Miles - Computerphile

https://www.youtube.com/watch?v=viJt_DXTfwA

Support us! https://www.patreon.com/mlst

MLST Discord: https://discord.gg/aNPkGUQtc5

YT: https://youtu.be/Vbi288CKgis

Michael Levin is a Distinguished Professor in the Biology department at Tufts University, and the holder of the Vannevar Bush endowed Chair. He is the Director of the Allen Discovery Center at Tufts and the Tufts Center for Regenerative and Developmental Biology. His research focuses on understanding the biophysical mechanisms of pattern regulation and harnessing endogenous bioelectric dynamics for rational control of growth and form.

The capacity to generate a complex, behaving organism from the single cell of a fertilized egg is one of the most amazing aspects of biology. Levin' lab integrates approaches from developmental biology, computer science, and cognitive science to investigate the emergence of form and function. Using biophysical and computational modeling approaches, they seek to understand the collective intelligence of cells, as they navigate physiological, transcriptional, morphognetic, and behavioral spaces. They develop conceptual frameworks for basal cognition and diverse intelligence, including synthetic organisms and AI.

Also joining us this evening is Irina Rish. Irina is a Full Professor at the Université de Montréal's Computer Science and Operations Research department, a core member of Mila - Quebec AI Institute, as well as the holder of the Canada CIFAR AI Chair and the Canadian Excellence Research Chair in Autonomous AI. She has a PhD in AI from UC Irvine. Her research focuses on machine learning, neural data analysis, neuroscience-inspired AI, continual lifelong learning, optimization algorithms, sparse modelling, probabilistic inference, dialog generation, biologically plausible reinforcement learning, and dynamical systems approaches to brain imaging analysis. 

Interviewer: Dr. Tim Scarfe

TOC:

[00:00:00] Introduction

[00:02:09] Emergence

[00:13:16] Scaling Laws

[00:23:12] Intelligence

[00:44:36] Transhumanism

Prof. Michael Levin

https://en.wikipedia.org/wiki/Michael_Levin_(biologist)

https://www.drmichaellevin.org/

https://twitter.com/drmichaellevin

Prof. Irina Rish

https://twitter.com/irinarish

https://irina-rish.com/

Dr. Patrick Lewis is a London-based AI and Natural Language Processing Research Scientist, working at co:here. Prior to this, Patrick worked as a research scientist at the Fundamental AI Research Lab (FAIR) at Meta AI. During his PhD, Patrick split his time between FAIR and University College London, working with Sebastian Riedel and Pontus Stenetorp. 

Patrick’s research focuses on the intersection of information retrieval techniques (IR) and large language models (LLMs). He has done extensive work on Retrieval-Augmented Language Models. His current focus is on building more powerful, efficient, robust, and update-able models that can perform well on a wide range of NLP tasks, but also excel on knowledge-intensive NLP tasks such as Question Answering and Fact Checking.

YT version: https://youtu.be/Dm5sfALoL1Y

MLST Discord: https://discord.gg/aNPkGUQtc5

Support us! https://www.patreon.com/mlst

References:

Patrick Lewis (Natural Language Processing Research Scientist @ co:here)

https://www.patricklewis.io/

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Patrick Lewis et al)

https://arxiv.org/abs/2005.11401

Atlas: Few-shot Learning with Retrieval Augmented Language Models (Gautier Izacard, Patrick Lewis, et al)

https://arxiv.org/abs/2208.03299

Improving language models by retrieving from trillions of tokens (RETRO) (Sebastian Borgeaud et al)

https://arxiv.org/abs/2112.04426

YT version (with references): https://www.youtube.com/watch?v=lxaTinmKxs0

Support us! https://www.patreon.com/mlst

MLST Discord: https://discord.gg/aNPkGUQtc5

Carla Cremer and Igor Krawczuk argue that AI risk should be understood as an old problem of politics, power and control with known solutions, and that threat models should be driven by empirical work. The interaction between FTX and the Effective Altruism community has sparked a lot of discussion about the dangers of optimization, and Carla's Vox article highlights the need for an institutional turn when taking on a responsibility like risk management for humanity.

Carla's “Democratizing Risk” paper found that certain types of risks fall through the cracks if they are just categorized into climate change or biological risks. Deliberative democracy has been found to be a better way to make decisions, and AI tools can be used to scale this type of democracy and be used for good, but the transparency of these algorithms to the citizens using the platform must be taken into consideration.

Aggregating people’s diverse ways of thinking about a problem and creating a risk-averse procedure gives a likely, highly probable outcome for having converged on the best policy. There needs to be a good reason to trust one organization with the risk management of humanity and all the different ways of thinking about risk must be taken into account. AI tools can help to scale this type of deliberative democracy, but the transparency of these algorithms must be taken into consideration.

The ambition of the EA community and Altruism Inc. is to protect and do risk management for the whole of humanity and this requires an institutional turn in order to do it effectively. The dangers of optimization are real, and it is essential to ensure that the risk management of humanity is done properly and ethically. By understanding the importance of aggregating people’s diverse ways of thinking about a problem, and creating a risk-averse procedure, it is possible to create a likely, highly probable outcome for having converged on the best policy.

Carla Zoe Cremer

https://carlacremer.github.io/

Igor Krawczuk

https://krawczuk.eu/

Interviewer: Dr. Tim Scarfe

TOC:

[00:00:00] Introduction: Vox article and effective altruism / FTX

[00:11:12] Luciano Floridi on Governance and Risk

[00:15:50] Connor Leahy on alignment

[00:21:08] Ethan Caballero on scaling

[00:23:23] Alignment, Values and politics

[00:30:50] Singularitarians vs AI-thiests

[00:41:56] Consequentialism

[00:46:44] Does scale make a difference?

[00:51:53] Carla's Democratising risk paper

[01:04:03] Vox article - How effective altruists ignored risk

[01:20:18] Does diversity breed complexity?

[01:29:50] Collective rationality

[01:35:16] Closing statements

Support us! https://www.patreon.com/mlst

MLST Discord: https://discord.gg/aNPkGUQtc5

YT version: https://youtu.be/YLNGvvgq3eg

We are living in an age of rapid technological advancement, and with this growth comes a digital divide. Professor Luciano Floridi of the Oxford Internet Institute / Oxford University believes that this divide not only affects our understanding of the implications of this new age, but also the organization of a fair society. 

The Information Revolution has been transforming the global economy, with the majority of global GDP now relying on intangible goods, such as information-related services. This in turn has led to the generation of immense amounts of data, more than humanity has ever seen in its history. With 95% of this data being generated by the current generation, Professor Floridi believes that we are becoming overwhelmed by this data, and that our agency as humans is being eroded as a result. 

According to Professor Floridi, the digital divide has caused a lack of balance between technological growth and our understanding of this growth. He believes that the infosphere is becoming polluted and the manifold of the infosphere is increasingly determined by technology and AI. Identifying, anticipating and resolving these problems has become essential, and Professor Floridi has dedicated his research to the Philosophy of Information, Philosophy of Technology and Digital Ethics. 

We must equip ourselves with a viable philosophy of information to help us better understand and address the risks of this new information age. Professor Floridi is leading the charge, and his research on Digital Ethics, the Philosophy of Information and the Philosophy of Technology is helping us to better anticipate, identify and resolve problems caused by the digital divide.

TOC:

[00:00:00] Introduction to Luciano and his ideas

[00:14:00] Chat GPT / language models

[00:28:45] AI risk / "Singularitarians" 

[00:37:15] Forms of governance

[00:43:56] Re-ontologising the world

[00:55:56] It from bit and Computationalism and philosophy without purpose

[01:03:05] Getting into Digital Ethics

Interviewer: Dr. Tim Scarfe

References:

GPT‐3: Its Nature, Scope, Limits, and Consequences [Floridi]

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3827044

Ultraintelligent Machines, Singularity, and Other Sci-fi Distractions about AI [Floridi]

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4222347

The Philosophy of Information [Floridi]

https://www.amazon.co.uk/Philosophy-Information-Luciano-Floridi/dp/0199232393

Information: A Very Short Introduction [Floridi]

https://www.amazon.co.uk/Information-Very-Short-Introduction-Introductions/dp/0199551375

https://en.wikipedia.org/wiki/Luciano_Floridi

https://www.philosophyofinformation.net/

Support us! https://www.patreon.com/mlst

MLST Discord: https://discord.gg/aNPkGUQtc5

YT version: https://youtu.be/YLNGvvgq3eg

(If music annoying, skip to main interview @ 14:14)

We are living in an age of rapid technological advancement, and with this growth comes a digital divide. Professor Luciano Floridi of the Oxford Internet Institute / Oxford University believes that this divide not only affects our understanding of the implications of this new age, but also the organization of a fair society. 

The Information Revolution has been transforming the global economy, with the majority of global GDP now relying on intangible goods, such as information-related services. This in turn has led to the generation of immense amounts of data, more than humanity has ever seen in its history. With 95% of this data being generated by the current generation, Professor Floridi believes that we are becoming overwhelmed by this data, and that our agency as humans is being eroded as a result. 

According to Professor Floridi, the digital divide has caused a lack of balance between technological growth and our understanding of this growth. He believes that the infosphere is becoming polluted and the manifold of the infosphere is increasingly determined by technology and AI. Identifying, anticipating and resolving these problems has become essential, and Professor Floridi has dedicated his research to the Philosophy of Information, Philosophy of Technology and Digital Ethics. 

We must equip ourselves with a viable philosophy of information to help us better understand and address the risks of this new information age. Professor Floridi is leading the charge, and his research on Digital Ethics, the Philosophy of Information and the Philosophy of Technology is helping us to better anticipate, identify and resolve problems caused by the digital divide.

TOC:

[00:00:00] Introduction to Luciano and his ideas

[00:14:40] Chat GPT / language models

[00:29:24] AI risk / "Singularitarians" 

[00:30:34] Re-ontologising the world

[00:56:35] It from bit and Computationalism and philosophy without purpose

[01:03:43] Getting into Digital Ethics

References:

GPT‐3: Its Nature, Scope, Limits, and Consequences [Floridi]

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3827044

Ultraintelligent Machines, Singularity, and Other Sci-fi Distractions about AI [Floridi]

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4222347

The Philosophy of Information [Floridi]

https://www.amazon.co.uk/Philosophy-Information-Luciano-Floridi/dp/0199232393

Information: A Very Short Introduction [Floridi]

https://www.amazon.co.uk/Information-Very-Short-Introduction-Introductions/dp/0199551375

https://en.wikipedia.org/wiki/Luciano_Floridi

https://www.philosophyofinformation.net/

Research has shown that humans possess strong inductive biases which enable them to quickly learn and generalize. In order to instill these same useful human inductive biases into machines, a paper was presented by Sreejan Kumar at the NeurIPS conference which won the Outstanding Paper of the Year award. The paper is called Using Natural Language and Program Abstractions to Instill Human Inductive Biases in Machines.

This paper focuses on using a controlled stimulus space of two-dimensional binary grids to define the space of abstract concepts that humans have and a feedback loop of collaboration between humans and machines to understand the differences in human and machine inductive biases. 

It is important to make machines more human-like to collaborate with them and understand their behavior. Synthesised discrete programs running on a turing machine computational model instead of a neural network substrate offers promise for the future of artificial intelligence. Neural networks and program induction should both be explored to get a well-rounded view of intelligence which works in multiple domains, computational substrates and which can acquire a diverse set of capabilities.

Natural language understanding in models can also be improved by instilling human language biases and programs into AI models. Sreejan used an experimental framework consisting of two dual task distributions, one generated from human priors and one from machine priors, to understand the differences in human and machine inductive biases. Furthermore, he demonstrated that compressive abstractions can be used to capture the essential structure of the environment for more human-like behavior. This means that emergent language-based inductive priors can be distilled into artificial neural networks, and AI  models can be aligned to the us, world and indeed, our values.

Humans possess strong inductive biases which enable them to quickly learn to perform various tasks. This is in contrast to neural networks, which lack the same inductive biases and struggle to learn them empirically from observational data, thus, they have difficulty generalizing to novel environments due to their lack of prior knowledge. 

Sreejan's results showed that when guided with representations from language and programs, the meta-learning agent not only improved performance on task distributions humans are adept at, but also decreased performa on control task distributions where humans perform poorly. This indicates that the abstraction supported by these representations, in the substrate of language or indeed, a program, is key in the development of aligned artificial agents with human-like generalization, capabilities, aligned values and behaviour.

References

Using natural language and program abstractions to instill human inductive biases in machines [Kumar et al/NEURIPS]

https://openreview.net/pdf?id=buXZ7nIqiwE

Core Knowledge [Elizabeth S. Spelke / Harvard]

https://www.harvardlds.org/wp-content/uploads/2017/01/SpelkeKinzler07-1.pdf

The Debate Over Understanding in AI's Large Language Models [Melanie Mitchell]

https://arxiv.org/abs/2210.13966

On the Measure of Intelligence [Francois Chollet]

https://arxiv.org/abs/1911.01547

ARC challenge [Chollet]

https://github.com/fchollet/ARC

Pedro Domingos, Professor Emeritus of Computer Science and Engineering at the University of Washington, is renowned for his research in machine learning, particularly for his work on Markov logic networks that allow for uncertain inference. He is also the author of the acclaimed book "The Master Algorithm".

Panel: Dr. Tim Scarfe

TOC:

[00:00:00] Introduction

[00:01:34] Galaxtica / misinformation / gatekeeping

[00:12:31] Is there a master algorithm?

[00:16:29] Limits of our understanding 

[00:21:57] Intentionality, Agency, Creativity

[00:27:56] Compositionality 

[00:29:30] Digital Physics / It from bit / Wolfram 

[00:35:17] Alignment / Utility functions

[00:43:36] Meritocracy  

[00:45:53] Game theory 

[01:00:00] EA/consequentialism/Utility

[01:11:09] Emergence / relationalism 

[01:19:26] Markov logic 

[01:25:38] Moving away from anthropocentrism 

[01:28:57] Neurosymbolic / infinity / tensor algerbra

[01:53:45] Abstraction

[01:57:26] Symmetries / Geometric DL

[02:02:46] Bias variance trade off 

[02:05:49] What seen at neurips

[02:12:58] Chalmers talk on LLMs

[02:28:32] Definition of intelligence

[02:32:40] LLMs 

[02:35:14] On experts in different fields

[02:40:15] Back to intelligence

[02:41:37] Spline theory / extrapolation

YT version:  https://www.youtube.com/watch?v=C9BH3F2c0vQ

References;

The Master Algorithm [Domingos]

https://www.amazon.co.uk/s?k=master+algorithm&i=stripbooks&crid=3CJ67DCY96DE8&sprefix=master+algorith%2Cstripbooks%2C82&ref=nb_sb_noss_2

INFORMATION, PHYSICS, QUANTUM: THE SEARCH FOR LINKS [John Wheeler/It from Bit]

https://philpapers.org/archive/WHEIPQ.pdf

A New Kind Of Science [Wolfram]

https://www.amazon.co.uk/New-Kind-Science-Stephen-Wolfram/dp/1579550088

The Rationalist's Guide to the Galaxy: Superintelligent AI and the Geeks Who Are Trying to Save Humanity's Future [Tom Chivers]

https://www.amazon.co.uk/Does-Not-Hate-You-Superintelligence/dp/1474608795

The Status Game: On Social Position and How We Use It [Will Storr]

https://www.goodreads.com/book/show/60598238-the-status-game

Newcomb's paradox

https://en.wikipedia.org/wiki/Newcomb%27s_paradox

The Case for Strong Emergence [Sabine Hossenfelder]

https://philpapers.org/rec/HOSTCF-3

Markov Logic: An Interface Layer for Artificial Intelligence [Domingos]

https://www.morganclaypool.com/doi/abs/10.2200/S00206ED1V01Y200907AIM007

Note; Pedro discussed “Tensor Logic” - I was not able to find a reference

Neural Networks and the Chomsky Hierarchy [Grégoire Delétang/DeepMind]

https://arxiv.org/abs/2207.02098

Connectionism and Cognitive Architecture: A Critical Analysis [Jerry A. Fodor and Zenon W. Pylyshyn]

https://ruccs.rutgers.edu/images/personal-zenon-pylyshyn/proseminars/Proseminar13/ConnectionistArchitecture.pdf

Every Model Learned by Gradient Descent Is Approximately a Kernel Machine [Pedro Domingos]

https://arxiv.org/abs/2012.00152

A Path Towards Autonomous Machine Intelligence Version 0.9.2, 2022-06-27 [LeCun]

https://openreview.net/pdf?id=BZ5a1r-kVsf

Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges [Michael M. Bronstein, Joan Bruna, Taco Cohen, Petar Veličković]

https://arxiv.org/abs/2104.13478

The Algebraic Mind: Integrating Connectionism and Cognitive Science [Gary Marcus]

https://www.amazon.co.uk/Algebraic-Mind-Integrating-Connectionism-D

Canadian Excellence Research Chair in Autonomous AI. Irina holds an MSc and PhD in AI from the University of California, Irvine as well as an MSc in Applied Mathematics from the Moscow Gubkin Institute. Her research focuses on machine learning, neural data analysis, and neuroscience-inspired AI. In particular, she is exploring continual lifelong learning, optimization algorithms for deep neural networks, sparse modelling and probabilistic inference, dialog generation, biologically plausible reinforcement learning, and dynamical systems approaches to brain imaging analysis. Prof. Rish holds 64 patents, has published over 80 research papers, several book chapters, three edited books, and a monograph on Sparse Modelling. She has served as a Senior Area Chair for NeurIPS and ICML.   Irina's research is focussed on taking us closer to the holy grail of Artificial General Intelligence.  She continues to push the boundaries of machine learning, continually striving to make advancements in neuroscience-inspired AI.

In a conversation about artificial intelligence (AI), Irina and Tim discussed the idea of transhumanism and the potential for AI to improve human flourishing. Irina suggested that instead of looking at AI as something to be controlled and regulated, people should view it as a tool to augment human capabilities. She argued that attempting to create an AI that is smarter than humans is not the best approach, and that a hybrid of human and AI intelligence is much more beneficial. As an example, she mentioned how technology can be used as an extension of the human mind, to track mental states and improve self-understanding. Ultimately, Irina concluded that transhumanism is about having a symbiotic relationship with technology, which can have a positive effect on both parties.

Tim then discussed the contrasting types of intelligence and how this could lead to something interesting emerging from the combination. He brought up the Trolley Problem and how difficult moral quandaries could be programmed into an AI. Irina then referenced The Garden of Forking Paths, a story which explores the idea of how different paths in life can be taken and how decisions from the past can have an effect on the present.

To better understand AI and intelligence, Irina suggested looking at it from multiple perspectives and understanding the importance of complex systems science in programming and understanding dynamical systems. She discussed the work of Michael Levin, who is looking into reprogramming biological computers with chemical interventions, and Tim mentioned Alex Mordvinsev, who is looking into the self-healing and repair of these systems. Ultimately, Irina argued that the key to understanding AI and intelligence is to recognize the complexity of the systems and to create hybrid models of human and AI intelligence.

Find Irina;

https://mila.quebec/en/person/irina-rish/

https://twitter.com/irinarish

YT version: https://youtu.be/8-ilcF0R7mI 

MLST Discord: https://discord.gg/aNPkGUQtc5

References;

The Garden of Forking Paths: Jorge Luis Borges [Jorge Luis Borges]

https://www.amazon.co.uk/Garden-Forking-Paths-Penguin-Modern/dp/0241339057

The Brain from Inside Out [György Buzsáki]

https://www.amazon.co.uk/Brain-Inside-Out-Gy%C3%B6rgy-Buzs%C3%A1ki/dp/0190905387

Growing Isotropic Neural Cellular Automata [Alexander Mordvintsev]

https://arxiv.org/abs/2205.01681

The Extended Mind [Andy Clark and David Chalmers]

https://www.jstor.org/stable/3328150

The Gentle Seduction [Marc Stiegler]

https://www.amazon.co.uk/Gentle-Seduction-Marc-Stiegler/dp/0671698877

Support us! https://www.patreon.com/mlst

Alan Chan is a PhD student at Mila, the Montreal Institute for Learning Algorithms, supervised by Nicolas Le Roux. Before joining Mila, Alan was a Masters student at the Alberta Machine Intelligence Institute and the University of Alberta, where he worked with Martha White. Alan's expertise and research interests encompass value alignment and AI governance. He is currently exploring the measurement of harms from language models and the incentives that agents have to impact the world. Alan's research focuses on understanding and controlling the values expressed by machine learning models. His projects have examined the regulation of explainability in algorithmic systems, scoring rules for performative binary prediction, the effects of global exclusion in AI development, and the role of a graduate student in approaching ethical impacts in AI research. In addition, Alan has conducted research into inverse policy evaluation for value-based sequential decision-making, and the concept of "normal accidents" and AI systems. Alan's research is motivated by the need to align AI systems with human values, and his passion for scientific and governance work in this field. Alan's energy and enthusiasm for his field is infectious. 

This was a discussion at NeurIPS. It was in quite a loud environment so the audio quality could have been better. 

References:

The Rationalist's Guide to the Galaxy: Superintelligent AI and the Geeks Who Are Trying to Save Humanity's Future [Tim Chivers]

https://www.amazon.co.uk/Does-Not-Hate-You-Superintelligence/dp/1474608795

The implausibility of intelligence explosion [Chollet]

https://medium.com/@francois.chollet/the-impossibility-of-intelligence-explosion-5be4a9eda6ec

Superintelligence: Paths, Dangers, Strategies [Bostrom]

https://www.amazon.co.uk/Superintelligence-Dangers-Strategies-Nick-Bostrom/dp/0199678111

A Theory of Universal Artificial Intelligence based on Algorithmic Complexity [Hutter]

https://arxiv.org/abs/cs/0004001

YT version: https://youtu.be/XBMnOsv9_pk 

MLST Discord: https://discord.gg/aNPkGUQtc5 

Support us! https://www.patreon.com/mlst

Professor Murray Shanahan is a renowned researcher on sophisticated cognition and its implications for artificial intelligence. His 2016 article ‘Conscious Exotica’ explores the Space of Possible Minds, a concept first proposed by philosopher Aaron Sloman in 1984, which includes all the different forms of minds from those of other animals to those of artificial intelligence. Shanahan rejects the idea of an impenetrable realm of subjective experience and argues that the majority of the space of possible minds may be occupied by non-natural variants, such as the ‘conscious exotica’ of which he speaks.  In his paper ‘Talking About Large Language Models’, Shanahan discusses the capabilities and limitations of large language models (LLMs). He argues that prompt engineering is a key element for advanced AI systems, as it involves exploiting prompt prefixes to adjust LLMs to various tasks. However, Shanahan cautions against ascribing human-like characteristics to these systems, as they are fundamentally different and lack a shared comprehension with humans. Even though LLMs can be integrated into embodied systems, it does not mean that they possess human-like language abilities. Ultimately, Shanahan concludes that although LLMs are formidable and versatile, we must be wary of over-simplifying their capacities and limitations.

YT version: https://youtu.be/BqkWpP3uMMU

Full references on the YT description. 

[00:00:00] Introduction

[00:08:51] Consciousness and  Consciousness Exotica

[00:34:59] Slightly Consciousness LLMs

[00:38:05] Embodiment

[00:51:32] Symbol Grounding 

[00:54:13] Emergence

[00:57:09] Reasoning

[01:03:16] Intentional Stance

[01:07:06] Digression on Chomsky show and Andrew Lampinen

[01:10:31] Prompt Engineering

Find Murray online:

https://www.doc.ic.ac.uk/~mpsha/

https://twitter.com/mpshanahan?lang=en

https://scholar.google.co.uk/citations?user=00bnGpAAAAAJ&hl=en

MLST Discord: https://discord.gg/aNPkGUQtc5

Support us! https://www.patreon.com/mlst

Sara Hooker is an exceptionally talented and accomplished leader and research scientist in the field of machine learning. She is the founder of Cohere For AI, a non-profit research lab that seeks to solve complex machine learning problems. She is passionate about creating more points of entry into machine learning research and has dedicated her efforts to understanding how progress in this field can be translated into reliable and accessible machine learning in the real-world.

Sara is also the co-founder of the Trustworthy ML Initiative, a forum and seminar series related to Trustworthy ML. She is on the advisory board of Patterns and is an active member of the MLC research group, which has a focus on making participation in machine learning research more accessible.

Before starting Cohere For AI, Sara worked as a research scientist at Google Brain. She has written several influential research papers, including "The Hardware Lottery", "The Low-Resource Double Bind: An Empirical Study of Pruning for Low-Resource Machine Translation", "Moving Beyond “Algorithmic Bias is a Data Problem”" and "Characterizing and Mitigating Bias in Compact Models". 

In addition to her research work, Sara is also the founder of the local Bay Area non-profit Delta Analytics, which works with non-profits and communities all over the world to build technical capacity and empower others to use data. She regularly gives tutorials on machine learning fundamentals, interpretability, model compression and deep neural networks and is dedicated to collaborating with independent researchers around the world.

Sara Hooker is famous for writing a paper introducing the concept of the 'hardware lottery', in which the success of a research idea is determined not by its inherent superiority, but by its compatibility with available software and hardware. She argued that choices about software and hardware have had a substantial impact in deciding the outcomes of early computer science history, and that with the increasing heterogeneity of the hardware landscape, gains from advances in computing may become increasingly disparate. Sara proposed that an interim goal should be to create better feedback mechanisms for researchers to understand how their algorithms interact with the hardware they use. She suggested that domain-specific languages, auto-tuning of algorithmic parameters, and better profiling tools may help to alleviate this issue, as well as provide researchers with more informed opinions about how hardware and software should progress. Ultimately, Sara encouraged researchers to be mindful of the implications of the hardware lottery, as it could mean that progress on some research directions is further obstructed. If you want to learn more about that paper, watch our previous interview with Sara.

YT version: https://youtu.be/7oJui4eSCoY

MLST Discord: https://discord.gg/aNPkGUQtc5

TOC:

[00:00:00] Intro

[00:02:53] Interpretability / Fairness

[00:35:29] LLMs

Find Sara:

https://www.sarahooker.me/

https://twitter.com/sarahookr

Support us! https://www.patreon.com/mlst

Hattie Zhou, a PhD student at Université de Montréal and Mila, has set out to understand and explain the performance of modern neural networks, believing it a key factor in building better, more trusted models. Having previously worked as a data scientist at Uber, a private equity analyst at Radar Capital, and an economic consultant at Cornerstone Research, she has recently released a paper in collaboration with the Google Brain team, titled ‘Teaching Algorithmic Reasoning via In-context Learning’. In this work, Hattie identifies and examines four key stages for successfully teaching algorithmic reasoning to large language models (LLMs): formulating algorithms as skills, teaching multiple skills simultaneously, teaching how to combine skills, and teaching how to use skills as tools. Through the application of algorithmic prompting, Hattie has achieved remarkable results, with an order of magnitude error reduction on some tasks compared to the best available baselines. This breakthrough demonstrates algorithmic prompting’s viability as an approach for teaching algorithmic reasoning to LLMs, and may have implications for other tasks requiring similar reasoning capabilities.

TOC

[00:00:00] Hattie Zhou

[00:19:49] Markus Rabe [Google Brain]

Hattie's Twitter - https://twitter.com/oh_that_hat

Website - http://hattiezhou.com/

Teaching Algorithmic Reasoning via In-context Learning [Hattie Zhou, Azade Nova, Hugo Larochelle, Aaron Courville, Behnam Neyshabur, and Hanie Sedghi]

https://arxiv.org/pdf/2211.09066.pdf

Markus Rabe [Google Brain]:

https://twitter.com/markusnrabe

https://research.google/people/106335/

https://www.linkedin.com/in/markusnrabe

Autoformalization with Large Language Models [Albert Jiang Charles Edgar Staats Christian Szegedy Markus Rabe Mateja Jamnik Wenda Li Yuhuai Tony Wu]

https://research.google/pubs/pub51691/

Discord: https://discord.gg/aNPkGUQtc5

YT: https://youtu.be/80i6D2TJdQ4

Support us! https://www.patreon.com/mlst

(On the main version we released; the music was a tiny bit too loud in places, and some pieces had percussion which was a bit distracting -- here is a version with all music removed so you have the option! )

David Chalmers is a professor of philosophy and neural science at New York University, and an honorary professor of philosophy at the Australian National University. He is the co-director of the Center for Mind, Brain, and Consciousness, as well as the PhilPapers Foundation. His research focuses on the philosophy of mind, especially consciousness, and its connection to fields such as cognitive science, physics, and technology. He also investigates areas such as the philosophy of language, metaphysics, and epistemology. With his impressive breadth of knowledge and experience, David Chalmers is a leader in the philosophical community.

The central challenge for consciousness studies is to explain how something immaterial, subjective, and personal can arise out of something material, objective, and impersonal. This is illustrated by the example of a bat, whose sensory experience is much different from ours, making it difficult to imagine what it's like to be one. Thomas Nagel's "inconceivability argument" has its advantages and disadvantages, but ultimately it is impossible to solve the mind-body problem due to the subjective nature of experience. This is further explored by examining the concept of philosophical zombies, which are physically and behaviorally indistinguishable from conscious humans yet lack conscious experience. This has implications for the Hard Problem of Consciousness, which is the attempt to explain how mental states are linked to neurophysiological activity. The Chinese Room Argument is used as a thought experiment to explain why physicality may be insufficient to be the source of the subjective, coherent experience we call consciousness. Despite much debate, the Hard Problem of Consciousness remains unsolved. Chalmers has been working on a functional approach to decide whether large language models are, or could be conscious. 

Filmed at #neurips22

Discord: https://discord.gg/aNPkGUQtc5

Pod: https://anchor.fm/machinelearningstreettalk/episodes/90---Prof--DAVID-CHALMERS---Slightly-Conscious-LLMs-e1sej50

TOC;

[00:00:00] Introduction

[00:00:40] LLMs consciousness pitch

[00:06:33] Philosophical Zombies

[00:09:26] The hard problem of consciousness

[00:11:40] Nagal's bat and intelligibility 

[00:21:04] LLM intro clip from NeurIPS

[00:22:55] Connor Leahy on self-awareness in LLMs

[00:23:30] Sneak peek from unreleased show - could consciousness be a submodule?

[00:33:44] SeppH

[00:36:15] Tim interviews David at NeurIPS (functionalism / panpsychism / Searle)

[00:45:20] Peter Hase interviews Chalmers (focus on interpretability/safety)

Panel:

Dr. Tim Scarfe

Dr. Keith Duggar

Contact David;

https://mobile.twitter.com/davidchalmers42

https://consc.net/

References;

Could a Large Language Model Be Conscious? [Chalmers NeurIPS22 talk]

https://nips.cc/media/neurips-2022/Slides/55867.pdf

What Is It Like to Be a Bat? [Nagel]

https://warwick.ac.uk/fac/cross_fac/iatl/study/ugmodules/humananimalstudies/lectures/32/nagel_bat.pdf

Zombies

https://plato.stanford.edu/entries/zombies/

zombies on the web [Chalmers]

https://consc.net/zombies-on-the-web/

The hard problem of consciousness [Chalmers]

https://psycnet.apa.org/record/2007-00485-017

David Chalmers, "Are Large Language Models Sentient?" [NYU talk, same as at NeurIPS]

https://www.youtube.com/watch?v=-BcuCmf00_Y

Support us! https://www.patreon.com/mlst

Dr. Walid Saba recently reviewed the book Machines Will Never Rule The World, which argues that strong AI is impossible. He acknowledges the complexity of modeling mental processes and language, as well as interactive dialogues, and questions the authors' use of "never." Despite his skepticism, he is impressed with recent developments in large language models, though he questions the extent of their success.

We then discussed the successes of cognitive science. Walid believes that something has been achieved which many cognitive scientists would never accept, namely the ability to learn from data empirically. Keith agrees that this is a huge step, but notes that there is still much work to be done to get to the "other 5%" of accuracy. They both agree that the current models are too brittle and require much more data and parameters to get to the desired level of accuracy.

Walid then expresses admiration for deep learning systems' ability to learn non-trivial aspects of language from ingesting text only. He argues that this is an "existential proof" of language competency and that it would be impossible for a group of luminaries such as Montague, Marvin Minsky, John McCarthy, and a thousand other bright engineers to replicate the same level of competency as we have now with LLMs. He then discusses the problem of semantics and pragmatics, as well as symbol grounding, and expresses skepticism about grounded meaning and embodiment. He believes that artificial intelligence should be used to solve real-world problems which require human intelligence but not believe that robots should be built to understand love or other subjective feelings.

We discussed the unique properties of natural human language. Walid believes that the core unique property is the ability to do abductive reasoning, which is the process of reasoning to the best explanation or understanding. Keith adds that there are two types of abduction - one for generating hypotheses and one for justifying them. In both cases, abductive reasoning involves choosing from a set of plausible possibilities.

Finally, we discussed the book "Machines Will Never Rule The World" and its argument that the current mathematics and technology is not enough to model complex systems. Walid agrees with the book's argument but is still optimistic that a new mathematics can be discovered. Keith suggests the possibility of an AGI discovering the mathematics to create itself. They also discussed how the book could serve as a reminder to temper the hype surrounding AI and to focus on exploration, creativity, and daring ideas. Walid ended by stressing the importance of science, noting that engineers should play within the Venn diagrams drawn by scientists, rather than trying to hack their way through it.

Transcript: https://share.descript.com/view/BFQb5iaegJC

Discord: https://discord.gg/aNPkGUQtc5

YT: https://youtu.be/IMnWAuoucjo

TOC:

[00:00:00] Intro

[00:06:52] Walid's change of heart on DL/LLMs and on the skeptics like Gary Marcus

[00:22:52] Symbol Grounding

[00:32:26] On Montague

[00:40:41] On Abduction

[00:50:54] Language of thought

[00:56:08] Why machines will never rule the world book review

[01:20:06] Engineers should play in the scientists Venn Diagram!

Panel;

Dr. Tim Scarfe

Dr. Keith Duggar

Mark Mcguill

Yann LeCun is a French computer scientist known for his pioneering work on convolutional neural networks, optical character recognition and computer vision. He is a Silver Professor at New York University and Vice President, Chief AI Scientist at Meta. Along with Yoshua Bengio and Geoffrey Hinton, he was awarded the 2018 Turing Award for their work on deep learning, earning them the nickname of the "Godfathers of Deep Learning".

Dr. Randall Balestriero has been researching learnable signal processing since 2013, with a focus on learnable parametrized wavelets and deep wavelet transforms. His research has been used by NASA, leading to applications such as Marsquake detection. During his PhD at Rice University, Randall explored deep networks from a theoretical perspective and improved state-of-the-art methods such as batch-normalization and generative networks. Later, when joining Meta AI Research (FAIR) as a postdoc with Prof. Yann LeCun, Randall further broadened his research interests to include self-supervised learning and the biases emerging from data-augmentation and regularization, resulting in numerous publications.

Episode recorded live at NeurIPS. 

YT: https://youtu.be/9dLd6n9yT8U (references are there)

Support us! https://www.patreon.com/mlst 

Host: Dr. Tim Scarfe

TOC:

[00:00:00] LeCun interview

[00:18:25] Randall Balestriero interview (mostly on spectral SSL paper, first ref)

Dr. Petar Veličković  is a Staff Research Scientist at DeepMind, he has firmly established himself as one of the most significant up and coming researchers in the deep learning space. He invented Graph Attention Networks in 2017 and has been a leading light in the field ever since pioneering research in Graph Neural Networks, Geometric Deep Learning and also Neural Algorithmic reasoning. If you haven’t already, you should check out our video on the Geometric Deep learning blueprint, featuring Petar. I caught up with him last week at NeurIPS. In this show, from NeurIPS 2022 we discussed his recent work on category theory and graph neural networks.

https://petar-v.com/

https://twitter.com/PetarV_93/

TOC:

Categories  (Cats for AI) [00:00:00]

Reasoning [00:14:44]

Extrapolation [00:19:09]

Ishan Misra Skit [00:27:50]

Graphs (Expander Graph Propagation) [00:29:18]

YT: https://youtu.be/1lkdWduuN14

MLST Discord: https://discord.gg/V25vQeFwhS

Support us! https://www.patreon.com/mlst

References on YT description, lots of them! 

Host: Dr. Tim Scarfe

In this NeurIPSs interview, we speak with Laura Ruis about her research on the ability of language models to interpret language in context. She has designed a simple task to evaluate the performance of widely used state-of-the-art language models and has found that they struggle to make pragmatic inferences (implicatures). Tune in to learn more about her findings and what they mean for the future of conversational AI.

Laura Ruis

https://www.lauraruis.com/

https://twitter.com/LauraRuis

BLOOM

https://bigscience.huggingface.co/blog/bloom

Large language models are not zero-shot communicators [Laura Ruis, Akbir Khan, Stella Biderman, Sara Hooker, Tim Rocktäschel, Edward Grefenstette]

https://arxiv.org/abs/2210.14986

[Zhang et al] OPT: Open Pre-trained Transformer Language Models

https://arxiv.org/pdf/2205.01068.pdf

[Lampinen] Can language models handle recursively nested grammatical structures? A case study on comparing models and humans

https://arxiv.org/pdf/2210.15303.pdf

[Gary Marcus] Horse rides astronaut

https://garymarcus.substack.com/p/horse-rides-astronaut

[Gary Marcus] GPT-3, Bloviator: OpenAI’s language generator has no idea what it’s talking about

https://www.technologyreview.com/2020/08/22/1007539/gpt3-openai-language-generator-artificial-intelligence-ai-opinion/

[Bender et al] On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?

https://dl.acm.org/doi/10.1145/3442188.3445922 

[janus] Simulators (Less Wrong)

https://www.lesswrong.com/posts/vJFdjigzmcXMhNTsx/simulators

First in our unplugged series live from #NeurIPS2022

We discuss natural language understanding, symbol meaning and grounding and Chomsky with Dr. Andrew Lampinen from DeepMind. 

We recorded a LOT of material from NeurIPS, keep an eye out for the uploads. 

YT version: https://youtu.be/46A-BcBbMnA

References

[Paul Cisek] Beyond the computer metaphor: Behaviour as interaction

https://philpapers.org/rec/CISBTC

Linguistic Competence (Chomsky reference)

https://en.wikipedia.org/wiki/Linguistic_competence

[Andrew Lampinen] Can language models handle recursively nested grammatical structures? A case study on comparing models and humans

https://arxiv.org/abs/2210.15303

[Fodor et al] Connectionism and Cognitive Architecture: A Critical Analysis

https://ruccs.rutgers.edu/images/personal-zenon-pylyshyn/proseminars/Proseminar13/ConnectionistArchitecture.pdf

[Melanie Mitchell et al] The Debate Over Understanding in AI's Large Language Models

https://arxiv.org/abs/2210.13966

[Gary Marcus] GPT-3, Bloviator: OpenAI’s language generator has no idea what it’s talking about

https://www.technologyreview.com/2020/08/22/1007539/gpt3-openai-language-generator-artificial-intelligence-ai-opinion/

[Bender et al] On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?

https://dl.acm.org/doi/10.1145/3442188.3445922

[Adam Santoro, Andrew Lampinen et al] Symbolic Behaviour in Artificial Intelligence

https://arxiv.org/abs/2102.03406

[Ishita Dasgupta, Lampinen et al] Language models show human-like content effects on reasoning

https://arxiv.org/abs/2207.07051

REACT - Synergizing Reasoning and Acting in Language Models

https://arxiv.org/pdf/2210.03629.pdf

https://ai.googleblog.com/2022/11/react-synergizing-reasoning-and-acting.html

[Fabian Paischer] HELM - History Compression via Language Models in Reinforcement Learning

https://ml-jku.github.io/blog/2022/helm/

https://arxiv.org/abs/2205.12258

[Laura Ruis] Large language models are not zero-shot communicators

https://arxiv.org/pdf/2210.14986.pdf

[Kumar] Using natural language and program abstractions to instill human inductive biases in machines

https://arxiv.org/pdf/2205.11558.pdf

Juho Kim

https://juhokim.com/

AI Helps Ukraine - Charity Conference

A charity conference on AI to raise funds for medical and humanitarian aid for Ukraine

https://aihelpsukraine.cc/

YT version: https://youtu.be/LgwjcqhkOA4

Support us!

https://www.patreon.com/mlst 

Dr. Joscha Bach (born 1973 in Weimar, Germany) is a German artificial intelligence researcher and cognitive scientist focusing on cognitive architectures, mental representation, emotion, social modelling, and multi-agent systems. 

http://bach.ai/

https://twitter.com/plinz

TOC:

[00:00:00] Ukraine Charity Conference and NeurIPS 2022

[00:03:40] Theory of computation, Godel, Penrose

[00:11:44] Modelling physical reality

[00:15:19] Is our universe infinite?

[00:24:30] Large language models, and on DL / is Gary Marcus hitting a wall?

[00:45:17] Generative models / Codex / Language of thought

[00:58:46] Consciousness (with Friston references)

References:

Am I Self-Conscious? (Or Does Self-Organization Entail Self-Consciousness?) [Friston]

https://www.frontiersin.org/articles/10.3389/fpsyg.2018.00579/full

Impact of Pretraining Term Frequencies on Few-Shot Reasoning [Yasaman Razeghi]

https://arxiv.org/abs/2202.07206

Deep Learning Is Hitting a Wall [Gary Marcus]

https://nautil.us/deep-learning-is-hitting-a-wall-238440/

Turing machines

https://en.wikipedia.org/wiki/Turing_machine

Lambda Calculus

https://en.wikipedia.org/wiki/Lambda_calculus

Godel's incompletness theorem

https://en.wikipedia.org/wiki/G%C3%B6del%27s_incompleteness_theorems

Oracle machine

https://en.wikipedia.org/wiki/Oracle_machine

Support us (and please rate on podcast app)

https://www.patreon.com/mlst 

In this show tonight with Prof. Julian Togelius (NYU) and Prof. Ken Stanley we discuss open-endedness, AGI, game AI and reinforcement learning.  

[Prof Julian Togelius]

https://engineering.nyu.edu/faculty/julian-togelius

https://twitter.com/togelius

[Prof Ken Stanley]

https://www.cs.ucf.edu/~kstanley/

https://twitter.com/kenneth0stanley

TOC:

[00:00:00] Introduction

[00:01:07] AI and computer games

[00:12:23] Intelligence

[00:21:27] Intelligence Explosion

[00:25:37] What should we be aspiring towards?

[00:29:14] Should AI contribute to culture?

[00:32:12] On creativity and open-endedness

[00:36:11] RL overfitting

[00:44:02] Diversity preservation

[00:51:18] Empiricism vs rationalism , in gradient descent the data pushes you around

[00:55:49] Creativity and interestingness (does complexity / information increase)

[01:03:20] What does a population give us?

[01:05:58] Emergence / generalisation snobbery

References;

[Hutter/Legg] Universal Intelligence: A Definition of Machine Intelligence

https://arxiv.org/abs/0712.3329

https://en.wikipedia.org/wiki/Artificial_general_intelligence

https://en.wikipedia.org/wiki/I._J._Good

https://en.wikipedia.org/wiki/G%C3%B6del_machine

[Chollet] Impossibility of intelligence explosion

https://medium.com/@francois.chollet/the-impossibility-of-intelligence-explosion-5be4a9eda6ec

[Alex Irpan] - RL is hard

https://www.alexirpan.com/2018/02/14/rl-hard.html

https://nethackchallenge.com/

Map elites

https://arxiv.org/abs/1504.04909

Covariance Matrix Adaptation for the Rapid Illumination of Behavior Space

https://arxiv.org/abs/1912.02400

[Stanley] - Why greatness cannot be planned

https://www.amazon.com/Why-Greatness-Cannot-Planned-Objective/dp/3319155237

[Lehman/Stanley] Abandoning Objectives: Evolution through the Search for Novelty Alone

https://www.cs.swarthmore.edu/~meeden/DevelopmentalRobotics/lehman_ecj11.pdf

We had a conversation with Aidan Gomez, the CEO of language-based AI platform Cohere. Cohere is a startup which uses artificial intelligence to help users build the next generation of language-based applications. It's headquartered in Toronto. The company has raised $175 million in funding so far.

Language may well become a key new substrate for software building, both in its representation and how we build the software. It may democratise software building so that more people can build software, and we can build new types of software. Aidan and I discuss this in detail in this episode of MLST.

Check out Cohere -- https://dashboard.cohere.ai/welcome/register?utm_source=influencer&utm_medium=social&utm_campaign=mlst

Support us!

https://www.patreon.com/mlst 

YT version: https://youtu.be/ooBt_di8DLs

TOC:

[00:00:00] Aidan Gomez intro

[00:02:12] What's it like being a CEO?

[00:02:52] Transformers

[00:09:33] Deepmind Chomsky Hierarchy

[00:14:58] Cohere roadmap

[00:18:18] Friction using LLMs for startups

[00:25:31] How different from OpenAI / GPT-3

[00:29:31] Engineering questions on Cohere

[00:35:13] Francois Chollet says that LLMs are like databases

[00:38:34] Next frontier of language models

[00:42:04] Different modes of understanding in LLMs

[00:47:04] LLMs are the new extended mind

[00:50:03] Is language the next interface, and why might that be bad?

References:

[Balestriero] Spine theory of NNs

https://proceedings.mlr.press/v80/balestriero18b/balestriero18b.pdf

[Delétang et al] Neural Networks and the Chomsky Hierarchy

https://arxiv.org/abs/2207.02098

[Fodor, Pylyshyn] Connectionism and Cognitive Architecture: A Critical Analysis

https://ruccs.rutgers.edu/images/personal-zenon-pylyshyn/docs/jaf.pdf

[Chalmers, Clark] The extended mind

https://icds.uoregon.edu/wp-content/uploads/2014/06/Clark-and-Chalmers-The-Extended-Mind.pdf

[Melanie Mitchell et al] The Debate Over Understanding in AI's Large Language Models

https://arxiv.org/abs/2210.13966

[Jay Alammar]

Illustrated stable diffusion

https://jalammar.github.io/illustrated-stable-diffusion/

Illustrated transformer

https://jalammar.github.io/illustrated-transformer/

https://www.youtube.com/channel/UCmOwsoHty5PrmE-3QhUBfPQ

[Sandra Kublik] (works at Cohere!)

https://www.youtube.com/channel/UCjG6QzmabZrBEeGh3vi-wDQ

This video is demonetised on music copyright so we would appreciate support on our Patreon! https://www.patreon.com/mlst 

We would also appreciate it if you rated us on your podcast platform. 

YT: https://youtu.be/_KVAzAzO5HU

Panel: Dr. Tim Scarfe, Dr. Keith Duggar

Guests: Prof. J. Mark Bishop, Francois Chollet, Prof. David Chalmers, Dr. Joscha Bach, Prof. Karl Friston, Alexander Mattick, Sam Roffey

The Chinese Room Argument was first proposed by philosopher John Searle in 1980. It is an argument against the possibility of artificial intelligence (AI) – that is, the idea that a machine could ever be truly intelligent, as opposed to just imitating intelligence.

The argument goes like this:

Imagine a room in which a person sits at a desk, with a book of rules in front of them. This person does not understand Chinese.

Someone outside the room passes a piece of paper through a slot in the door. On this paper is a Chinese character. The person in the room consults the book of rules and, following these rules, writes down another Chinese character and passes it back out through the slot.

To someone outside the room, it appears that the person in the room is engaging in a conversation in Chinese. In reality, they have no idea what they are doing – they are just following the rules in the book.

The Chinese Room Argument is an argument against the idea that a machine could ever be truly intelligent. It is based on the idea that intelligence requires understanding, and that following rules is not the same as understanding.

in this detailed investigation into the Chinese Room, Consciousness and Syntax vs Semantics, we interview luminaries J.Mark Bishop and Francois Chollet and use unreleased footage from our interviews with David Chalmers, Joscha Bach and Karl Friston. We also cover material from Walid Saba and interview Alex Mattick from Yannic's Discord. 

This is probably my favourite ever episode of MLST. I hope you enjoy it!  With Keith Duggar. 

Note that we are using clips from our unreleased interviews from David Chalmers and Joscha Bach -- we will release those shows properly in the coming weeks. We apologise for delay releasing our backlog, we have been busy building a startup company in the background.

TOC: 

[00:00:00] Kick off

[00:00:46] Searle

[00:05:09] Bishop introduces CRA

[00:00:00] Stevan Hardad take on CRA 

[00:14:03] Francois Chollet dissects CRA

[00:34:16] Chalmers on consciousness

[00:36:27] Joscha Bach on consciousness

[00:42:01] Bishop introduction

[00:51:51] Karl Friston on consciousness

[00:55:19] Bishop on consciousness and comments on Chalmers 

[01:21:37] Private language games (including clip with Sam Roffey)

[01:27:27] Dr. Walid Saba on the chinese room (gofai/systematicity take)

[00:34:36] Bishop: on agency / teleology

[01:36:38] Bishop: back to CRA

[01:40:53] Noam Chomsky on mysteries 

[01:45:56] Eric Curiel on math does not represent

[01:48:14] Alexander Mattick on syntax vs semantics

Thanks to: Mark MC on Discord for stimulating conversation, Alexander Mattick, Dr. Keith Duggar, Sam Roffey. Sam's YouTube channel is https://www.youtube.com/channel/UCjRNMsglFYFwNsnOWIOgt1Q

Vitaliy Chiley  is a Machine Learning Research Engineer at the next-generation computing hardware company Cerebras Systems. We spoke about how DL workloads including sparse workloads can run faster on Cerebras hardware.

[00:00:00] Housekeeping

[00:01:08] Preamble

[00:01:50] Vitaliy Chiley Introduction

[00:03:11] Cerebrus architecture

[00:08:12] Memory management and FLOP utilisation

[00:18:01] Centralised vs decentralised compute architecture

[00:21:12] Sparsity

[00:23:47] Does Sparse NN imply Heterogeneous compute?

[00:29:21] Cost of distributed memory stores?

[00:31:01] Activation vs weight sparsity

[00:37:52] What constitutes a dead weight to be pruned?

[00:39:02] Is it still a saving if we have to choose between weight and activation sparsity?

[00:41:02] Cerebras is a cool place to work

[00:44:05] What is sparsity? Why do we need to start dense? 

[00:46:36] Evolutionary algorithms on Cerebras?

[00:47:57] How can we start sparse? Google RIGL

[00