Former DeepMind researcher publishes article arguing LLMs don't truly reason
Understand the story
On October 2, 2026, MIT Technology Review reported that Thore Graepel, a former core member of the DeepMind AlphaGo team, wrote in an article that large language models are essentially still fast associative next-token prediction, that their chain-of-thought is not an independent reasoning mechanism, and that they lack reasoning capability in the scientific sense. This view continues the earlier controversy over whether LLMs truly possess reasoning ability. Graepel's article has so far sparked industry discussion on the nature of LLM reasoning, with no public responses from other parties reported yet.
AI overview of reporting · Updated 16 days ago
Reporting timeline
Explore different aspects of the story through its reporting.
- MIT Technology Review · AIDon’t be fooled—LLMs don’t reason
Thore Graepel, a former core member of DeepMind's AlphaGo team, argues in an essay that large language models are essentially fast-associative next-token prediction, that their chains of thought are not an independent reasoning mechanism, and that they do not reason in the scientific sense.