Shortly after SpaceX’s IPO, AI researcher Yann LeCun has come out with pointed criticism of Elon Musk’s AI company xAI. In an interview with US broadcaster CNBC, conducted on the sidelines of the VivaTech technology conference in Paris – the city that is also home to LeCun’s own company, AMI Labs – he described xAI as “kind of a failure.” He said he does not expect the company to be able to keep pace with OpenAI and Anthropic at the frontier of AI development.

LeCun, who was previously Meta’s chief AI scientist and is often referred to as one of the “godfathers of AI” because of his early work in the field, based his assessment primarily on staff departures. Over the past year, several of xAI’s co-founders have left the organisation. As a result, he argued, Musk is now in a “very, very difficult” position when it comes to recruiting top AI talent – not least because, in LeCun’s view, Musk has not behaved well towards the previous team.

The remarks add to a long-running feud between the two men. LeCun and Musk have clashed publicly for years, on topics ranging from AI to what LeCun has described as the Tesla CEO’s “conspiracy theories” on social media. Musk, for his part, has accused LeCun of having been out of touch with AI for a long time.

SpaceX merger and the Memphis data centres

In February, Musk merged SpaceX with xAI – a deal that valued the company at 1.25 trillion US dollars. In the first quarter (ended 31 March), SpaceX’s AI segment, which includes xAI, posted a 2.5-billion-dollar operating loss according to the available figures.

LeCun also pointed to xAI’s “huge infrastructure” – a reference to the Colossus 1 and Colossus 2 data centres in Memphis, Tennessee. xAI rents this capacity out to other companies in order to recoup its costs; both Google and Anthropic have rented computing capacity there, he noted. “I’m not very positive about the prospect of xAI,” LeCun said.

His own company, AMI Labs, closed a one-billion-dollar funding round in March, at a pre-money valuation of 3.5 billion dollars. AMI Labs is working on so-called world models, which LeCun sees as key to the next stage of AI development.

Beyond xAI, LeCun expressed broader scepticism about the economic sustainability of the industry. AI is proving more expensive than expected; OpenAI CEO Sam Altman recently called the costs a “huge issue.” While the prices of AI services are rising, he said, the cost of running them is not falling fast enough. Providers are losing money, and usage is essentially funded by investors – “that can’t go on for very long,” LeCun said.

Labs such as OpenAI and Anthropic will therefore have to raise prices or cut costs, he argued – otherwise the industry faces a “big bubble explosion.”

Criticism of large language models

LeCun is a prominent critic of the limitations of large language models (LLMs), the foundation of the current generation of leading AI products. Instead, he favours world models, which take a different approach by seeking to build an understanding of how the real or simulated world works – including objects, cause and effect, and actions.

“I personally don’t think we’re going to have generalised, reliable agentic systems until they’re based on world models,” LeCun said. LLMs, he noted, are useful for areas such as coding or maths. But the cost of running such systems at this level of performance is very high relative to what users are willing to pay.

Aus Datenschutz-Gründen ist dieser Inhalt ausgeblendet. Die Einbettung von externen Inhalten kann in den Datenschutz-Einstellungen aktiviert werden: