Ep 248: AI and Philosophy of Science

whatshot 37 snips

Oct 13, 2025

The discussion dives into the misconceptions around AI and knowledge creation, challenging the predictive doom narratives that dominate current discourse. It critiques the reliance on Bayesian reasoning and highlights the philosophical blind spots in understanding AI's potential. The talk emphasizes that science isn't just data extrapolation, urging a reevaluation of personhood and creativity in the context of technology. It sheds light on the pitfalls of both AI doomers and accelerationists, advocating for a more nuanced understanding of how knowledge and science evolve.

01:25:39

forum

Ask episode

web_stories

AI Snips

view_agenda

Chapters

menu_book

Books

auto_awesome

Transcript

info_circle

Episode notes

insights

INSIGHT

Knowledge Creation Is Unpredictable

Knowledge creation is inherently unpredictable and cannot be reduced to data extrapolation.
Brett Hall argues that failing to recognise literal knowledge creation leads to wrong forecasts about AI's future.

insights

INSIGHT

AI Doom Is A New Priestly Prophecy

Hall frames AI debates as a clash of new priesthoods who prophesy futures people crave to hear.
He warns prophetic, sensational claims (e.g., 'If Anyone Builds It, Everyone Dies') distort policy and public perception.

question_answer

ANECDOTE

Early Admiration For Celebrity Physicists

Brett Hall recalls watching Paul Davies on TV and believing physicists had special insight into big questions.
That early admiration shaped his initial desire to study cosmology and find ultimate answers.

Get the Snipd Podcast app to discover more snips from this episode

Why Superintelligence Arguments Felt Wrong

00:54 • 2min

chevron_right

Preparing a Response to AI Doomism

02:57 • 46sec

chevron_right

Why Philosophy of Science Matters for AI

03:43 • 2min

chevron_right

The New Priesthood: Physicists and Technologists

05:36 • 2min

chevron_right

Personhood and the Limits of Physics

07:48 • 2min

chevron_right

Big Questions Aren't Solved by Credentials

09:54 • 3min

chevron_right

Life, Consciousness and the Missing Explanations

13:00 • 4min

chevron_right

The New Priesthoods and Prophecy About the Future

17:00 • 2min

chevron_right

Knowledge Creation Is Inherently Unpredictable

18:40 • 2min

chevron_right

Early Critique of Bostrom and Bayesianism

20:19 • 5min

chevron_right

Why Doomers Recycle the Same Arguments?

25:23 • 5min

chevron_right

Publicity, Levity, and the Politics of Doom

30:14 • 6min

chevron_right

Is Science Just Extrapolating Data?

35:49 • 3min

chevron_right

Explanations Versus Data Extrapolation

38:37 • 34sec

chevron_right

Doom Scenarios Depend on a False Model of Science

39:11 • 3min

chevron_right

How Does Explanatory Knowledge Appear?

41:43 • 5min

chevron_right

Survey of Philosophy of Science Positions

47:11 • 3min

chevron_right

Popper, Deutsch and the Quest to Explain Creativity

49:59 • 50sec

chevron_right

Transformers, Attention, and Limits of LLMs

50:49 • 3min

chevron_right

Models, Understanding, and Scientific Insight

53:22 • 6min

chevron_right

Could an LLM Solve Quantum Gravity?

58:54 • 2min

chevron_right

Why There Can Be No Reliable Test for AGI

01:01:08 • 3min

chevron_right

Doomers vs. Accelerationists: Both Misguided

01:04:00 • 4min

chevron_right

Are Humans Just Prediction Engines?

01:07:53 • 4min

chevron_right

Bad Philosophy Stymies Knowledge Growth

Will AI Make Students Stupid?

01:17:12 • 2min

chevron_right

Are Current Gen AI Systems Truly Creative?

01:18:53 • 4min

chevron_right

Practical Benefits and Real Risks of AI Tools

More Big Questions with Paul Davies

• Mentioned in 3 episodes

Science in a Free Society

Paul Feyerabend

#114

• Mentioned in 118 episodes

The Beginning of Infinity

Explanations That Transform the World

David Deutsch

In this groundbreaking book, David Deutsch argues that explanations have a fundamental place in the universe and that improving them is the basic regulating principle of all successful human endeavor. The book takes readers on a journey through various fields of science, history of civilization, art, moral values, and the theory of political institutions. Deutsch explains how we form new explanations and drop bad ones, and discusses the conditions under which progress, which he argues is potentially boundless, can and cannot happen. He emphasizes the importance of good explanations, which he defines as those that are 'hard to vary' and have 'reach', and argues that these explanations are central to the Enlightenment way of thinking and to all scientific and philosophical progress.

#720

• Mentioned in 33 episodes

A New Kind Of Science

Stephen Wolfram

In 'A New Kind of Science,' Stephen Wolfram presents a comprehensive study of computational systems, particularly cellular automata, and argues that the study of simple programs can revolutionize various fields of science. The book, which took over a decade to complete, introduces the Principle of Computational Equivalence and the concept of computational irreducibility. Wolfram demonstrates how simple rules can generate complex behavior, similar to patterns observed in nature, and discusses the implications of these findings for fields such as physics, biology, and mathematics. The book is known for its extensive use of computer graphics and its attempt to establish a new foundational science based on computational principles.

#3342

• Mentioned in 10 episodes

The end of science

Facing the Limits of Knowledge in the Twilight of the Scientific Age

John Horgan

In 'The End of Science,' John Horgan discusses the idea that science may be nearing its theoretical limits. He interviews various eminent scholars, including Fred Hoyle, Noam Chomsky, and John Wheeler, to explore topics such as God, superstrings, quarks, consciousness, and the Omega Point. Horgan argues that science is facing significant challenges, including the constraints imposed by special relativity, quantum mechanics, and chaos theory. He also critiques the trend of 'ironic science,' where theoreticians engage in speculative and often untestable theories, potentially reducing science to a status akin to literary criticism. The book presents a contrarian view on the future of scientific discovery and its place in the hierarchy of disciplines.

• Mentioned in 42 episodes

The Fabric of Reality

The Science of Parallel Universes--and Its Implications

David Deutsch

In 'The Fabric of Reality', David Deutsch explores a four-strand theory of everything, integrating Hugh Everett's many-worlds interpretation of quantum physics, Karl Popper's epistemology, Alan Turing's theory of computation, and Richard Dawkins's evolutionary theory. The book discusses the implications of these theories, including the concept of the multiverse, quantum computers, time travel, and the ultimate fate of the universe. Deutsch argues that these theories, when taken literally and jointly, reveal a unified, objective, and comprehensible fabric of reality.

#467

• Mentioned in 44 episodes

If Anyone Builds It, Everyone Dies

Eliezer Yudkowsky

Nate Soares

This book delves into the potential risks of advanced artificial intelligence, arguing that the development of superintelligence could lead to catastrophic consequences for humanity. The authors present a compelling case for the need for careful consideration and regulation of AI development. They explore various scenarios and potential outcomes, emphasizing the urgency of addressing the challenges posed by rapidly advancing AI capabilities. The book is written in an accessible style, making complex ideas understandable to a broad audience. It serves as a call to action, urging policymakers and researchers to prioritize AI safety and prevent potential existential threats.

#9624

• Mentioned in 4 episodes

The big questions

Steven Landsburg

#94

• Mentioned in 133 episodes

Superintelligence

Paths, Dangers, Strategies

Nick Bostrom

In this book, Nick Bostrom delves into the implications of creating superintelligence, which could surpass human intelligence in all domains. He discusses the potential dangers, such as the loss of human control over such powerful entities, and presents various strategies to ensure that superintelligences align with human values. The book examines the 'AI control problem' and the need to endow future machine intelligence with positive values to prevent existential risks.

This is the extended "director's cut" of a talk delivered for "RatFest 2025" (next year to be "Conjecture Con"). This also serves as a supplement to my "Doom Debates" interview which can be found here: https://youtu.be/koubXR0YL4A?si=483M6SPOKwbQYmzb

It is simply assumed some version of "Bayesian reasoning" is how AI will "create" knowledge. This misconception permeates the https://ai-2027.com paper, as well as Bostrom and Yudkowsky's work on this, as well as that of every other AI "Doomer" and even on the other extreme the so-called "AI-Accelerationists". All of that indicates a deep misconception about how new explanations are generated which comes from a deep misconception about how science works because almost no one in the field of AI seems to think the *philosophy of* science is even relevant. I explain what has gone wrong:

00:00 Introduction

09:14 The Big Questions and the new Priesthoods

18:40 Nick Bostrom and Superintelligence

25:10 If anyone builds it, everyone dies and Yudkowsky.

33:32 Prophecy, Inevitability, Induction and Bayesianism.

41:42 Popper, Kuhn, Feyerabend and Lakatos.

49:40 AI researchers ignore The Philosophy of Science.

58:46 A new test for AGI from Sam Altman and David Deutsch?

1:03:35 Accelerationists, Doomers and “Everyone dies”.

1:10:21 Conclusions

1:15:35 Audience Questions

Home Top podcasts Popular guests Top books