Sceptical investors worry whether advances in AI will make money

Gordon Ritter, founder of San Francisco-based venture fund Emergence Capital, believes that recent developments in the field of artificial intelligence represent a significant technological advance.

He just cannot see a way to make money out of them.

“Everyone has stars in their eyes about what could happen,” says Ritter, whose firm was an early investor in successful start ups such as Zoom. “There’s a flow [of opinion that AI] will do everything. We’re going against that flow.”

The scepticism reflects a tension among Silicon Valley VCs, who are caught between excitement over AI and a broader tech downturn that has led to falling investment in start-ups over the past year.

But the recent launch of “generative AI” tools such as OpenAI’s ChatGPT chatbot, capable of answering complex questions with text in natural-sounding language, has resulted in fresh excitement over the potential emergence of a new group of industry-defining companies.

Last week, the Financial Times revealed that Andreessen Horowitz led an investment of more than $200mn into Character.ai, valuing the chatbot-maker at about $1bn. Meanwhile, one-year-old start up Inflection AI, founded by DeepMind’s co-founder Mustafa Suleyman and LinkedIn creator Reid Hoffman, is in talks to raise up to $675mn, having already raised $225mn last year.

In January, Microsoft confirmed a “multibillion-dollar investment” in ChatGPT-maker OpenAI. People familiar with the talks previously said OpenAI was seeking $10bn from Microsoft at a $29bn valuation.

Yet many VCs express caution, put off not only by eye-watering valuations, but also the huge amount of capital AI groups require as they build “foundation models” — machine-learning systems that require huge amounts of data and computing power to operate.

One investor said that, because of the huge amount of capital and computing resources required, recent leaps in generative AI were comparable to landing on the moon: a massively impressive technical achievement, only replicable by those with nation-state level wealth.

“Companies are extremely overvalued and the only justifiable investment thesis is to get in incredibly early,” said another veteran investor. “Otherwise you’re only buying in because of FOMO.”

Microsoft’s deal with OpenAI ensures the company can access the tech giant’s Azure cloud computing platform. Earlier this month, Google paid about $300mn to take a 10 per cent stake in Anthropic, at a time the AI start-up was buying substantial computing resources from the search company’s cloud computing division.

Enthusiasm for AI comes after many top investors, such as Sequoia Capital, have also been burnt after pouring cash into nascent crypto groups, only for cryptocurrency values to fall over the past year.

The value of deals struck by US venture funds halved between the first and fourth quarters of last year, from $81bn to $41bn, according to PitchBook. Crypto deals plunged further, down more than 80 per cent over the same period.

AI’s potential has drawn in the likes of Sarah Guo who, a year ago, led investment into the crypto sector for venture capital firm Greylock, having also been an angel investor into cryptocurrency exchange FTX.

FTX has since collapsed into bankruptcy, but Guo has raised more than $100mn to invest into artificial intelligence with her new fund Conviction. She argues now is a good time to invest and competition for AI deals is fierce but admits “some of these valuations are nonsense”.

Priced out of companies building foundation models, many investors are looking instead to the applications which will be built on top of them — just as huge businesses have been built from the mobile apps that operate on mobile devices.

That has led some to bide their time until they sense the arrival of a “killer app” with clear commercial potential. “If ChatGPT is the iPhone, we’re seeing a lot of calculator apps,” said Christina Melas-Kyriazi, a partner at Bain Capital Ventures. “We’re looking for Uber.”

The buzz around generative AI has increased as consumers have been able to play with the technology, conversing with ChatGPT to write speeches, or using text-to-image programs such as Dall-E to create fantastical drawings.

As more people begin using the programs, the prevalence of errors and undesirable responses has also grown. Inaccuracies are deemed unacceptable in sensitive sectors such as health and defence, meaning a fundamentally disruptive application could be a long way off.

“AI makes interesting mistakes that humans don’t; you have to think very carefully before using that in defence,” said Yisroel Brumer, president of venture capital fund Red Cell Partners. “The first time you enter a combat situation you will see something you’ve never seen before.”

Recent progress in generative AI, he added, was “incredibly impressive and well beyond where we estimated we’d be today” but “life and death decisions make it very difficult to delegate to a system that makes mistakes that are hard to understand and hard to predict.”

Other investors warn that those pivoting from crypto deals to AI investments may be swapping one flashy technology with few commercial applications with another.

“Sometimes the bubble moves from one place to the other, the money has to go somewhere,” said one venture capital investor who bet big on crypto. “During these hype cycles there will be many things which are overvalued. There are many toys.”