BEHIND THE AI BUBBLE, ANOTHER TECH REVOLUTION COULD BE BREWING / THE FINANCIAL TIMES OP EDITORIAL

Behind the AI bubble, another tech revolution could be brewing

Today’s eye-popping valuations are based on the assumption that LLMs are the only game in town

Gillian Tett

© Efi Chalikopoulou

This week Jensen Huang, the charismatic leather jacket-wearing head of Nvidia, has dominated headlines. 

No wonder: on Wednesday his company, which is a bellwether for the artificial intelligence sector, revealed third-quarter revenues of $57bn, 62 per cent higher year on year.

That sent tech stocks soaring, after they had slumped earlier on fears that AI valuations are in bubble territory. 

In other words, they are displaying “elements of irrationality”, to cite Sundar Pichai, head of Google owner Alphabet.

And Huang duly celebrated. 

“There has been a lot of talk about an AI bubble. 

[But] from our vantage point we see something very different,” he declared, noting sky-high demand for his chips.

Phew — or so some investors might respond. 

But while Huang hogged the spotlight, a second, less visible, figure also created news this week: Yann LeCun, the French-US scientist who has won multiple prizes as an AI pioneer.

On Wednesday he confirmed the FT’s reporting that he will soon leave his current post as Meta’s chief scientist to create his own start-up. 

This is standard fare for Silicon Valley’s febrile culture, and tech insiders have expected his departure ever since Mark Zuckerberg, Meta’s founder, appointed Alexandr Wang, 28, to lead his new “superintelligence” team. 

Asking LeCun, 65, to report to a member of Gen Z was a stretch.

But what is more notable are LeCun’s plans. 

In recent years, the AI race has been dominated by the technical approach based on large language models and statistics outlined in a revolutionary 2017 paper. 

So-called “transformer” tech has since unleashed products like ChatGPT.

But LeCun thinks LLMs are reaching their limits. 

Instead, as he noted in a 2022 paper, he now favours “world models”, which use an approach to sorting information that is inspired by how humans learn. 

“LLMs are great, they’re useful, we should invest in them — a lot of people are going to use them,” he said this month. 

“[But] they are not a path to human-level intelligence . . . so for the next revolution, we need to take a step back.”

Does this matter? 

Some techies think not. 

After all, innovation never occurs in a straight line. 

And right now, as Huang notes, AI enthusiasm is so high that demand for Nvidia’s products seems almost bottomless. 

But the symbolism of Huang and LeCun’s near-simultaneous announcements is delicious — and should not be ignored. 

Until now, the main issue that has sparked the “AI-as-bubble” fears is that the corporate demand will be slower and/or lower than optimists expect, meaning that revenue projections are too optimistic. 

This risk is real, given that surveys show a wide variation in whether companies can reap productivity gains from AI — or not.

But LeCun is one symbol of a second risk. 

Today’s eye-popping AI valuations are partly based on the assumption that LLMs are the main game in town — and can only be exploited by the current capex and capital-heavy approach that Big Tech is unleashing.

But when the Chinese company DeepSeek released its models earlier this year, it showed there are ways to build cheaper, scaled-down variants of AI, raising the prospect that LLMs will become commoditised. 

And LeCun is not the only player who thinks current LLMs might be supplanted.

The tech behemoth IBM says it is developing variants of so-called neuro-symbolic AI. 

“By augmenting and combining the strengths of statistical AI, like machine learning, with the capabilities of humanlike symbolic knowledge and reasoning, we’re aiming to create a revolution in AI, rather than an evolution,” it explains.

Chinese and western researchers are also exploring variants of neuro-symbolic AI while Fei-Fei Li, the so-called “Godmother of AI”, is developing a world model version called “spatial intelligence”.

None of these alternatives seems ready to fly right now; indeed LeCun acknowledges huge practical impediments to his dream. 

But if they do ever work, it would raise many questions.

Might the oodles of current capex by Big Tech turn out to be wasted, and become stranded assets? 

Could the collateral underpinning the associated soaring debt need to be depreciated? 

Jeff Bezos, Amazon founder, says this is a “good bubble” since it will leave behind useful infrastructure. 

But AI chips have less shelf life than the tracks installed in the 19th-century railway mania, or fibre optic cables in the dotcom bubble.

Might a new AI approach break the control of the Big Tech companies currently trying to outspend their way to LLM dominance, crushing the minnows? 

Could AI development eventually echo the pattern seen when VHS video replaced Betamax, or Facebook beat MySpace? 

I don’t know. 

And maybe the Big Tech leaders will just adapt to — and keep dominating — whatever emerges. 

But DeepSeek was a warning shot, and sometimes it only takes a small event to create a tipping point in sentiment. 

So investors should marvel at Huang this week; but they should also watch the hordes of other less visible AI luminaries. 

This (very real) revolution is still at an early stage.