Arc
Abstraction and Reasoning Corpus (for AGI), a benchmark of unique, never-before-seen puzzles designed to test adaptation to novelty rather than memorization; easy for humans, very hard for LLMs.
As defined in 280 | François Chollet on Deep Learning and the Meaning of Intelligence
Other definitions
A gene (A-R-C) essential for memory in mice and people, whose mutations are associated with dementia and schizophrenia; its protein forms virus-like capsules and it originated as a domesticated ancient virus.
88 | Neil Shubin on Evolution, Genes, and Dramatic Transitions
Where it comes up · 4
- 280 | François Chollet on Deep Learning and the Meaning of Intelligence4 claims
- Dario Amodei (Anthropic CEO) — The hidden pattern behind every AI breakthrough1 claim
- 88 | Neil Shubin on Evolution, Genes, and Dramatic Transitions1 claim
- How China Trade Policy Reshapes U.S. Dollar Dominance | Lyn Alden1 claim
Most important claims · 7 of 7Search all claims for “Arc” →
- The Arc gene, essential for memory in mice and people (its mutation associated with dementia and schizophrenia), produces protein spherules that look like the HIV capsid; Jason Shepherd showed Arc is actually an ancient virus that invaded a distant ancestor and, instead of being knocked out, was domesticated and repurposed—the capsule that protects viral genetic information cell-to-cell now carries genetic material neuron-to-neuron to make memories.88 | Neil Shubin on Evolution, Genes, and Dramatic Transitions
- ARC (Abstraction and Reasoning Corpus for AGI), released in 2019, is a benchmark of unique, original puzzles never seen before and partly kept private, designed so that memorization cannot help; humans solve them easily while deep learning and LLMs score very poorly—GPT-3 scored zero in the 2020 Kaggle competition—because each puzzle is genuinely novel and LLMs have essentially no ability to adapt to such novelty.280 | François Chollet on Deep Learning and the Meaning of Intelligence
- General intelligence is the ability to synthesize on the fly a new correct program matching a novel problem from just a few examples, rather than fetching or interpolating memorized programs; if a system could do this efficiently it would have AGI and would also be able to solve ARC, whereas LLMs fail precisely because all they can do is fetch.280 | François Chollet on Deep Learning and the Meaning of Intelligence
- An LLM cannot bootstrap a smarter AI because it is constrained by its training data to interpolating known code snippets; even an average senior software engineer—a tremendously capable novel problem-solver—cannot invent AGI, so an LLM with no novel problem-solving ability certainly cannot, and cannot even solve a trivial ARC problem that a four-year-old can.280 | François Chollet on Deep Learning and the Meaning of Intelligence
- To avoid a 'one-shot' takeover risk when testing whether a powerful model is dangerous, Anthropic plans to set capability thresholds well below dangerous levels — testing for prerequisites to survival in the wild like opening an AWS account and making money for itself — and proceed with more rigorous, careful tests as capabilities increase, extrapolating with factors of two of compute.Dario Amodei (Anthropic CEO) — The hidden pattern behind every AI breakthrough
- Chollet is relaunching the ARC competition on Kaggle after four years with over one million dollars in prizes, aiming to reach roughly human-level (85%) performance, explicitly to incentivize people to invent new approaches beyond LLMs—which he characterizes as an off-ramp rather than a path to AGI.280 | François Chollet on Deep Learning and the Meaning of Intelligence
- The host and speaker observed rapid development in Bitcoin second-layer solutions (lightning, eCash, Liquid, Arc) with increasing interoperability, creating new applications like Hashpool (mining pool decentralization via eCash+Lightning) that would have been impossible previously.How China Trade Policy Reshapes U.S. Dollar Dominance | Lyn Alden