The biggest figures in artificial intelligence sparred over the dangers of the rapidly advancing technology at the World Economic Forum this week, as hype swirled around a $500bn AI infrastructure project touted by Donald Trump.
AI pioneers including Google DeepMind chief Sir Demis Hassabis, Anthropic co-founder Dario Amodei and “godfather of AI” computer scientist Yoshua Bengio used the gathering in Davos to reiterate stark warnings about the AI threats, as commercial interests and geopolitical rivalries steamroller concerns about safety.
While Hassabis acknowledged that the “genie can’t be put back in the bottle”, he said artificial general intelligence — when computers surpass human cognitive capabilities — could threaten civilisation if it runs out of control or is hijacked by bad actors. This is particularly the case with large language models that are “open source” and accessible by all.
“There’s much more at stake here than just companies or products,” the Nobel Prize winner said in an interview with the Financial Times. “[It’s] the future of humanity, the human condition and where we want to go as a society.”
Amodei, whose start-up makes the chatbot Claude and is backed by Google and Amazon, said he was concerned about authoritarian governments using AI and was “very worried about 1984 scenarios, or worse”.
“Science doesn’t know how we can control machines that are even at our level of intelligence, and even worse if they’re smarter than us,” added Bengio during a panel. “There are people who are saying, ‘Don’t worry, we’ll figure it out.’ But if we don’t figure it out, do you understand the consequences?”
Their stance was criticised as hypocritical by Yann LeCun, chief AI scientist at Meta, which has spent billions developing an open source LLM called Llama. He said that such concerns were belied by his rivals’ fierce competition to build, and sell, the best models.
“Yoshua and Dario have made opinions against open source and that’s actually very dangerous,” he said in an interview. “Obstacles to open source distribution would lead to regulatory capture by a few players, either of the west coast of the US or China . . . [putting] power in the hands of a small number of people.
“It’s very strange from people like Dario. We met yesterday where he said that the benefits and risks of AI are roughly on the same order of magnitude, and I said, ‘if you really believe this, why do you keep working on AI?’” LeCun added. “So I think he is a little two-faced on this.”
Whilst scientists and engineers debated the risk-reward of AI, business executives showed unfettered enthusiasm for the technology.
“There are no contrarians,” said Ervin Tu, president of Dutch tech investment group Prosus. “If you have any appreciation for what large language models and agents trained on them can do, you would be hard-pressed as a human not to conclude that they are transformational and will be incredibly disruptive in every industry.”
On Wednesday, the febrile atmosphere was further charged by OpenAI, SoftBank and Oracle announcing a $500bn US AI infrastructure joint venture called “Stargate”.
Trump hosted their chief executives, Sam Altman, Masayoshi Son and Larry Ellison, in the Oval Office on Tuesday, before signing executive orders this week that would eliminate many guardrails around the development of the technology. The new US president said the moves would ensure American primacy in the technology.
“At OpenAI, we believe infrastructure is destiny,” said OpenAI chief financial officer Sarah Friar. “[Stargate] is about more compute. More compute builds better models. Better models answer more complex problems and deliver more benefits for people and businesses.”
Stargate dominated debate in Davos for the rest of the week, with many including Elon Musk taking to his social networking site X to question how the trio would fund the vast expenditure promised.
The FT reported on Friday that Stargate has not yet secured the funding it requires, will receive no government financing and will serve only OpenAI once completed. So far, SoftBank and OpenAI intend to put forward more than $15bn each for the project, hoping to raise a combination of equity from their existing backers and debt to fund Stargate.
The new venture was also taken as the latest evidence of a fissure in the relationship between Altman and Microsoft chief executive Satya Nadella and his top AI executive Mustafa Suleyman, the former DeepMind cofounder who left his own startup and joined Microsoft early last year.
“The tensions that surfaced between Mustafa Suleyman and Sam Altman at Davos last year were just the beginning,” said Salesforce chief executive Marc Benioff, which competes with Microsoft in selling AI-powered agents to businesses.
“Microsoft is now accelerating its own AI development . . . This pattern reflects Microsoft’s history with its ‘partners,’” Benioff added. “This could mark the beginning of the end for the relationship, making it critical for OpenAI to expand to other platforms quickly.”
“Marc has no idea what he’s talking about,” said Microsoft spokesperson Frank Shaw.
Microsoft has invested almost $14bn in OpenAI since 2019 and in return negotiated rights to its intellectual property and to be its exclusive cloud computing provider. But the latter agreement was terminated alongside the announcement of Stargate.
In Davos, Nadella also cast doubt on the Stargate spending pledges and touted Microsoft’s planned $80bn in capital expenditure.
“All I know is I’m good for my $80bn,” he said, later replying to Musk on social media platform X: “And all this money is not about hyping AI, but is about building useful things for the real world!”
Stargate is just the latest example of an infrastructure arms race for data centres in the US as it prepares for the next leg of the AI economic boom. Musk’s xAI built a supercomputer called “Colossus” containing 100,000 interconnected Nvidia chips in just three months last year and has pledged to expand the number 10-fold.
BlackRock and Microsoft are preparing to launch a $30bn AI investment fund to build data centres and energy projects to meet growing demands stemming from the tech sector. On Friday, Meta chief Mark Zuckerberg said the company would spend between $60bn-$65bn on capital infrastructure this year while expanding its AI teams.
“I’ve had nonstop customer meetings, across every sector. I don’t think there’s a single CEO I’ve spoken to who doesn’t know they need to be deploying AI,” said OpenAI’s Friar. “AI isn’t just on the agenda; it is the agenda. It is no longer just an abstract concept or futuristic vision. It’s here.”
Additional reporting by Harriet Agnew in Davos