It looks like it could be the end of the AI hype cycle

• Some AI leaders are starting to question whether they should be buying the hype.
• AI startups raised around $50 billion in 2023 as some predicted AI to be as important as the internet.
• But few companies have managed to build profitable businesses, and talk of AGI can feel farfetched.

Some AI leaders are starting to ask themselves the question: should we be buying the hype?

Now is probably a good time to start reflecting on where the industry is. Almost a year and a half on from the launch of ChatGPT, the hype around the technology is seemingly everywhere.

Luminaries like Bill Gates have said that AI will be as consequential as mobile phones and the internet. Elon Musk, meanwhile, has suggested that it will fast-track the development of technology that surpasses human intelligence by the end of the decade.

Others have touted AI's coming impact on the economy, with apps built on top of large language models from OpenAI, Google, and Meta being marketed as superchargers to worker productivity across industries.

In February, for instance, Cathie Wood's ARK Invest released an eyebrow-raising report that predicted AI would boost total GDP by 130% between 2023 and 2030. It charted that forecast next to the GDP impact of the steam engine in the 19th century which it pegged at around 30%.

Investors have taken note too. Generative AI and AI-related startups raised almost $50 billion last year, according to Crunchbase. That included Microsoft's massive funding round into OpenAI as well as huge deals for new startups like France's Mistral.

For some, though, it's becoming evident that this hype is getting way out of hand.

'The entire industry is based on hype'

One person reckoning with the hype is Gary Marcus.

Last weekend, the AI expert and New York University professor — who testified at a Senate hearing on AI alongside OpenAI CEO Sam Altman last year — predicted the generative AI bubble could burst within the next 12 months.

"The entire industry is based on hype," Marcus wrote in a Substack blog published on Sunday. It's a bold claim but it rests on some recent developments that are tough to ignore.

For one, as Marcus noted, the industry is spending much more money than it's raking in, raising questions about how long that can last.

He noted a recent Wall Street Journal report, which cited a presentation from venture capital firm Sequoia, that claimed $50 billion had been spent by the industry in 2023 on Nvidia chips that provide vital computing power for AI, but just $3 billion was made in revenue.

"That's obviously not sustainable," he wrote.

Some might contend that this concern is only real to those with a short-termism approach to value creation, who can only see as far as the next quarter.

That may be a fair contention. Nvidia just introduced the world to its new Blackwell GPUs, which the industry will once again be ready to invest heavily in, given the important role they are set to play in helping train the next generation of large language models.

But the verdict is still out on whether those behind foundation AI models dependent on expensive chips can turn them into viable, profitable businesses. And last month offered several reasons to raise doubt.

Inflection AI saw its boss, Mustafa Suleyman, jump ship to Microsoft along with a bunch of its researchers, while seemingly de-prioritizing the consumer focus of its chatbot Pi. AI image generator Stability AI also lost its chief, Emad Mostaque, amid reports of financial pressure.

It's hard not to argue that these businesses grew off the back of hype. Inflection AI grew rapidly to a $4 billion valuation after earning the favor of brand-name backers like Microsoft, Nvidia, and ex-Google CEO Eric Schmidt; Stability AI did similarly in securing a $1 billion valuation.

But that hype might be a big problem itself — something Google DeepMind boss Demis Hassabis acknowledged in an interview with the Financial Times published Sunday.

He said the huge amounts of money flowing into AI startups "brings with it a whole attendant bunch of hype and maybe some grifting" as "we're talking about all sorts of things that are just not real."

Though there have been lots of very real developments in AI since ChatGPT's launch, a lot of hype has surrounded developments that are not, such as artificial general intelligence — a hypothetical version of AI that puts its intelligence on par with humans.

AGI's arrival is imminent if you're to believe some, like Musk. Others though, like Google deep learning expert François Chollet, are keen to stress just how far off human intelligence today's AI models actually are.

Until some real signs of money-making, human-level intelligence do emerge, expect more to call it all hype.