Recommended Free Tools
François Chollet’s five-to-ten-year figure was an opinion about how research priorities and openness affect progress toward artificial general intelligence (AGI), not a measured delay. In a June 11, 2024 interview with Dwarkesh Patel, Chollet argued that OpenAI helped reduce publication of frontier research and that enthusiasm for large language models (LLMs) drew researchers and resources toward one approach. The headline compresses his fuller wording: he said progress “towards AGI” had been set back “by quite a few years, probably like 5-10 years.”
What did François Chollet say?
In the June 11, 2024 conversation with host Dwarkesh Patel, Chollet said: “OpenAI basically set back progress towards AGI by quite a few years, probably like 5-10 years.” The popular headline version shortens that sentence, so it should not be treated as a verbatim transcript.
As an Amazon Associate I earn from qualifying purchases.
Chollet was making a judgment about the direction of AI research. He did not cite a calculation showing that AGI had been delayed by a specific number of years, and the interview does not establish five to ten years as a verified statistic or settled forecast.
Why did Chollet think progress had been set back?
He connected his estimate to two changes he believed had narrowed the field of research. These were his criticisms, not consensus findings.
#1 Best Overall
Less publication of frontier research
Chollet argued that OpenAI helped usher in an environment where less frontier research was published openly. In his view, reduced sharing makes it harder for researchers to examine, reproduce, and build on one another’s work.
Research concentrated on LLMs
He also argued that hype around LLMs drew researchers and resources toward that approach, leaving less attention for alternatives. Chollet described LLMs as an “off-ramp” on the path to AGI. That phrase captures his position; it does not establish that language models cannot contribute to future AI capabilities.
Rank #2
Why did ARC come up in the conversation?
The interview was framed around ARC and the launch of the ARC-AGI Prize. ARC stands for Abstraction and Reasoning Corpus, a benchmark designed around novel problems involving visual, spatial, and counting concepts.
In a typical ARC task, a solver sees example input and output grids, infers the transformation connecting them, then applies that rule to a new grid. The aim is to test adaptation to unfamiliar problems rather than straightforward recall of training examples. Chollet’s view was that ARC requires a solver to “try new ideas.”
During the 2024 conversation, he discussed discrete program search and program synthesis as approaches that were working well in this context. The discussion contrasted search over a smaller set of program primitives with methods that use learned building blocks and recombine them. Chollet treated these as different points on a spectrum, not as a simple consumer-style choice between products.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Does the five-to-ten-year estimate have independent confirmation?
No independent, dated statistic in the cited sources quantifies how many years OpenAI’s actions delayed AGI. Big Think’s August 10, 2024 explainer also attributes a shortened version of the claim to Chollet, but that is confirmation of the quote’s circulation, not evidence that the estimate is measurable or established.
The ARC Prize Foundation currently describes its work as advancing open-source AGI research through benchmarks and prizes. Its definition of AGI centers on a system matching human learning efficiency; that is the foundation’s framing, not a universal definition. Its present-day activities provide context for the ARC discussion, but they do not validate Chollet’s estimate.
Quick Recap
Best Value
How should readers interpret the quote?
- It is an attributed opinion: Chollet offered a causal argument about publication practices and research priorities.
- The number is not a measurement: Five to ten years was his qualified estimate, not a result from a study.
- ARC is an example in the debate: Its novel grid puzzles illustrate the kind of adaptation Chollet wanted AI systems to demonstrate; the interview does not settle whether LLMs can achieve that or what counts as AGI.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
