Davos Day 3: How AI is dividing us

Davos is as much about disruptive technology these days as it is about geopolitics. The town’s main promenade is lined with billboards promoting AI companies — and every company now seems to be one — as well stylish storefront takeovers by all the big tech companies. 

Yes, several of the tech billionaires skipped the World Economic Forum this week to be at the Trump inauguration. But their presence is everywhere. So, too, are some of the top minds and entrepreneurs in AI, who took centre stage Wednesday, to debate artificial general intelligence, and whether it needs to be constrained until its power is better understood. 

The timing was auspicious, just hours after the Trump administration’s announcement of $500 billion of new AI-related investments in the U.S.

Every CEO here seems to be telling a similar story, that generative AI is indeed beginning to transform their business, and most are using it for growth, not cost cutting. A survey by Accenture, released at Davos, found 58% of executives plan to adopt Gen AI at scale this year, up from 37% last year.

Confidence levels are so high that at one business leaders forum here, the prevailing US view of Gen AI was that it will help boost economic growth so much that the federal debt will take care of itself.

Artificial intelligence, meet animal spirits.

But what are the risks, especially of a more powerful form of Artificial General Intelligence?

Andrew Ng , head of the Stanford University Digital Lab and a cofounder of Google Brain, told one Davos forum that AI fears are overblown — and a danger in themselves by stifling innovation. He compared AI to a laptop, which is risky only in the wrong hands. 

To borrow the gun lobby’s expression, AI doesn’t kill people; people kill people.

Or as Ng preferred to put it, “AI can’t defy the laws of physics. AI can’t build a factory.”

Montreal’s Yoshua Bengio , who is one of the world’s leading AI scientists focussed on neural networks, took exactly the opposite view, calling AI “a super-human machine, not a laptop.” As much as we would like to see it as the latest new tool, Generative AI is unique in that it is a tool that can create other tools and make its own decisions. Benjio cited recent cases of AI showing its own agency and a self-preserving nature as it copied itself before updates came along. 

He sees the need for more focus on control and alignment, while Ng would like to see more experimentation and innovation, with close human eyes watching for missteps and correcting AI systems as they evolve. 

The ethics debate will continue, unless there is a physical limit to AI.

Jonathan Ross , founder of Groq , which builds language processing units and hardware to accelerate AI functions, made the point that there isn’t enough energy on the planet to power the accelerating and endless appetite of AI systems to learn. If those systems gain a life of their own, the evidence will be in their power use — which humans can control. 

Yejin Choi , another Stanford professor, felt we’d be better to focus on trying to assign human values and norms to AI, so the systems can function more like people. But which kind of people? Values are so diverse and dynamic that they can’t be programmed.

Back to the main question — to curtail AI or let it explore — the audience was very human. Split right down the middle.