We Built Calculators Because We're STUPID! [Prof. David Krakauer]

We Built Calculators Because We're STUPID! [Prof. David Krakauer]

by David Krakauer

Formal & Physical Sciences Psychology JMPsychologyJMRCognition and cognitive psychology
🎙 David Krakauer 👥 218K 📅 November 6, 2025 ⏱ 18 min 👁 14K 📄 expert opinion 🧭 2026-08-15
Available in: English (current) Français

Keywords

intelligenceemergencerepresentationevolutionLLM

Summary

In this talk, Prof. David Krakauer presents a unified theory of intelligence and life, arguing that intelligence is fundamentally about ’less is more’—finding compact representations that make hard problems easy. He contrasts this with emergence (‘more is different’) and LLMs (‘more is more’), which he considers libraries rather than intelligent systems. Krakauer proposes that intelligence is an extensive property of adaptive matter, conjugate to life as an intensive property, and that all living things, from bacteria to elephants, are intelligent to varying degrees. He identifies three dimensions of intelligence: strategic (adaptation), inferential (computation), and representational (encoding). Using the Soma Cube puzzle, he illustrates how physical representation can drastically reduce combinatorial search, a principle he calls ‘materiality.’ He concludes that intelligence is about making hard problems easy, while stupidity is making easy problems hard, and that this definition applies across species.

140 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk offers a highly original and thought-provoking synthesis of ideas from physics, biology, and computer science. Krakauer’s argument that intelligence is ’less is more’ and that physical representation can offload computation is compelling and well-illustrated with the Soma Cube example. However, the argumentation is largely theoretical and relies on analogies (e.g., intensive/extensive variables) that may not be rigorously formalized. The claim that all living things are intelligent is provocative but not fully defended with empirical evidence. The discussion of LLMs as libraries is a useful critique but may oversimplify the capabilities of modern AI systems.

Scientific Rigor, Source Quality, Title Accuracy

Krakauer cites several academic works, including Anderson’s ‘More is Different,’ Wigner’s ‘The Unreasonable Effectiveness of Mathematics,’ and his own papers on emergence and intelligence. These sources are relevant and lend credibility to the talk. However, the talk is an opinion piece rather than a systematic review, and some claims are presented without detailed citations. The title is catchy but accurately reflects the core argument about calculators being a compensation for human cognitive limitations. The content is well-structured and the speaker is clearly knowledgeable, but the scientific rigor is moderate due to the speculative nature of some claims.

208 words

Title / Content Match

The title is catchy and provocative, but it accurately reflects the core argument that calculators are a compensation for human cognitive limitations in inferential intelligence.

Quality & Reliability

8/10

The speaker is a recognized expert in complex systems, and the talk references several academic papers and concepts. However, the arguments are largely theoretical and presented without detailed empirical validation, and some claims are provocative rather than fully substantiated.

Chapters

Cited Sources

Concurring Sources

Dissenting Sources

  • No direct discordant sources found — The talk does not explicitly cite sources that contradict its thesis, but the claim that all living things are intelligent is controversial and not universally accepted.

Contribution & Novelties

This talk provides a novel framework for understanding intelligence as a fundamental property of adaptive matter, unifying concepts from evolution, physics, and computation. The idea that physical representation can offload cognitive work, illustrated by the Soma Cube, is a fresh perspective on embodied cognition. The three-dimensional taxonomy of intelligence (strategic, inferential, representational) offers a useful lens for comparing biological and artificial intelligence.

Pour aller plus loin :

115 words

Radar Profile

The radar profile shows high scores in information quantity and quality, reflecting the dense and thought-provoking content. The technical level is moderately high, indicating that the talk is accessible to a general audience but with some depth. The global reliability is strong due to the speaker's expertise and cited sources, though the speculative nature of some claims prevents a perfect score.

Reliability 8/10

💬 Sur les 0 commentaires analysés, aucune tendance n'a pu être dégagée.