The AI bubble

Samuel Buchmann

24/5/2024

Translation: Patrik Stainbrook

The gold rush around artificial intelligence is unending. Meanwhile, the real benefits of this technology remain limited. If this doesn’t change, the bubble will soon burst.

Generative artificial intelligence (AI) is the biggest tech hype since the internet. The giants of Silicon Valley are in a breakneck race. OpenAI, Microsoft, Google and Meta are trying to outdo each other with slick chatbots. The promise? AI will increase our productivity immeasurably.

Maybe. Someday. But current expectations are anything but realistic.

Greed is getting out of hand

This can be seen on Wall Street. There, euphoria is running rampant. Investors are pumping huge amounts of capital into anything to do with AI – including Alphabet, Amazon, Meta, Microsoft and Nvidia. Between 1 January 2023 and 31 March 2024, around six trillion dollars flowed into the shares of these five companies.

This astronomical figure is almost seven times the gross domestic product of Switzerland (0.9 trillion) – and 22 per cent of that of the USA (27 trillion). As a result, the five companies increased their market capitalisation by 132 per cent. A large part of this increase is probably down to the high expectations of AI. The MSCI World global equity market index rose by only 31 per cent in the same period.

Jensen Huang is raking it in

Plenty of capital is also flowing between the big companies – especially towards Nvidia. Alphabet, Amazon, Meta and Microsoft are buying as many AI processors as they can get their hands on. With them, they’re building the huge servers needed to train AI models and process requests. Nvidia’s Tensor Core GPUs are so far ahead of the competition that the chip manufacturer has a virtual monopoly.

A Ponzi scheme based on hope

So far, Alphabet, Amazon, Meta and Microsoft can afford this. They’ve also increased their profits in recent years. At Meta, however, this is mainly due to cost-cutting measures and traditional income from advertising. It’s a similar story for Alphabet, Amazon and Microsoft. They also rent out cloud computing power to smaller companies that can’t get their own Nvidia GPUs.

AI remains unreliable

Whether this’ll pay off in the end remains to be seen. Last year, ChatGPT caused a stir across the world – the language model awakened sci-fi fears. But the bot wasn’t only impressive, it was also unreliable. It made things up, relied on stereotypes and spread misinformation – just like the subsequent chatbots from Google and Meta. This unreliability was dismissed as an early misstep.

But it’s remained to this day. Google’s Gemini and OpenAI’s GPT-4o are still often wrong. Every request is like a coin toss. Maybe the answer will be correct, but maybe not. Regardless of whether it’s software code, medical advice or stock market data.

Smoke grenades

Such features make headlines and suggest rapid development. But these are mostly smoke grenades to distract from the basic problem – AI models aren’t ready for the market because they just make too many mistakes. They’re like a bad intern who has to be constantly monitored.

Even if you disregard legal and ethical concerns, there will be a structural problem at some point. Google’s AI Overview could ensure that traffic stays with Google. This would cause the business model of many websites to disappear – and thus also the incentive to produce new content at all. But it’s precisely this content that provides the necessary nourishment for the data octopuses of current models. AI bites the hand that feeds it.

Wall Street’s confidence is fragile

Perhaps Silicon Valley will still find a solution. One approach, for example, could be specialised models trained with less but better data. Artificial intelligence is undoubtedly a useful technology. There are applications, and there will be many more that nobody has thought of yet. But the revolution may come more slowly and less comprehensively than the hype suggests.

For now, investors still have patience. They’re happy for Google and Microsoft to present new AI products every few months – regardless of whether they’re any good. Nobody knows how long this will last. But the moment of truth will come, and the bubble is full to bursting.

Pop.

Samuel Buchmann

Senior Editor

Samuel.Buchmann@digitecgalaxus.ch

My fingerprint often changes so drastically that my MacBook doesn't recognise it anymore. The reason? If I'm not clinging to a monitor or camera, I'm probably clinging to a rockface by the tips of my fingers.

---

Opinion

This is a subjective opinion of the editorial team. It doesn't necessarily reflect the position of the company.

Show all

These articles might also interest you