Axios House: Executives and academics agree there's collective benefit to diversifying AI

DAVOS, Switzerland — Artificial intelligence is in dire need of diversification when it comes to where the technology is located and who's designing it, industry leaders say.

Why it matters: A robust ecosystem of governments, organizations, stakeholders, and models can alleviate the risks of concentrated power.

Axios' Ina Fried spoke to Cohere co-founder and CEO Aidan Gomez and Stanford Institute for Human-Centered Artificial Intelligence (HAI) senior fellow Yejin Choi. The Jan. 19 conversations were sponsored by Turing.

What they're saying: "The past year has sort of woken everyone up to the reality that we're [all] going to have to carry our weight," Gomez told Fried.

• "We can't all depend on … a single liberal democracy," Gomez said.

AI shows up across sectors — communications, finance, energy, and beyond. Gomez, whose organization has offices in Seoul, Paris, Toronto, San Francisco, and other major cities around the world, describes the diversification of where the technology lives as a matter of national security.

• "Each country cannot depend on a single other country. They need to spread their bets and make sure that they're resilient in the event that one of those gets switched off."

Choi, a professor and computer scientist, said she is worried about the concentration of power by certain corporations.

• "The tricky thing about the current way we all rely on LLM APIs is that it's, by and large, created by a few very rich tech companies," Choi said.
• Those who currently stand to gain the most when it comes to AI are "well-intended people, at least at the beginning," Choi added, but "AI should benefit all of humanity, not just the few who are in power."

The bottom line: "AI should be: of human, for human, by human," Choi said.

Content from the sponsored segment:

In a View from the Top sponsored segment, Turing founder and CEO Jonathan Siddharth said that, in order for AI models to do their best work and deliver the best value, they have to "touch reality" and learn from real people.

• "It's only by deploying it incrementally, slowly, iteratively that we can make these systems safe and useful."

Go deeper: Watch the full interviews on YouTube.