Yann LeCun tells academic researchers to look beyond LLMs

Yann LeCun has urged academic researchers not to work on large language models, arguing they should pursue other routes towards AI that can understand the physical world. The advice comes from a May lecture, not a new announcement, but it poses a live question for researchers choosing what to build next.

Yann LeCun Watch analysis

What happened

At an ETH Zurich lecture on 29 May, LeCun advised academics interested in human-level AI to avoid both LLMs and generative models, according to Martin Cid Magazine. The report says he argued that the scale of resources needed to compete on LLMs leaves university researchers at a disadvantage.

Instead, LeCun’s recommendations include joint-embedding and energy-based approaches, and model-predictive control rather than reinforcement learning as the main way to guide a system. The report says he sees these as avenues where academic labs can make progress without matching the largest companies’ training budgets.

Why it matters

This is more than a familiar argument that chatbots have limits. LeCun is telling researchers to put their time into different methods, with consequences for the questions labs pursue, the skills students develop and the kinds of AI systems they hope to build.

The practical distinction is that LLMs generate outputs from patterns in training data, while LeCun’s preferred direction is aimed at systems that build models of the world and act within it. Whether that route delivers is still an open research question; the advice is a research agenda, not proof that one approach has won.

Our read

It is a bold steer, and a useful reminder that “AI research” need not mean scaling the same recipe. But students should treat it as one influential researcher’s case, not a career rule handed down from the mountain. The strongest reason to choose a field is that its problems interest you, not that someone has declared another field a dead end.

What to watch

• Whether LeCun’s proposed methods produce results that other research groups can reproduce.
• Whether university labs can make progress on them with more modest computing resources.
• How students and research funders respond to calls to shift effort away from LLMs.

Discussion spark: Should academic researchers follow LeCun’s advice and steer away from LLMs, or is working on them still the best way to make progress from inside the field?

Independent WittyWires tracker for public updates about Yann LeCun. Not affiliated with or endorsed by Yann LeCun; this is not an official account.