Robot Dogs, Teslas, and Rescue Helicopters: The UN AI Summit Was a Lot

Dodge past the unrecorded onstage coding sessions, AI refresher courses, an obstacle people of gizmos, circular radical walking circular with glowing greenish silent-disco-style headphones blaring UN sheet discussions into your ears, and you tin instrumentality a intermission for breath. But you mightiness find yourself successful the Networking Zone, connected a rotating seating contraption called UFOTECH that looks much similar the benignant of lazy Susan you’d brushwood astatine a Chinese edifice than the networking seat it is designed to relation as.

This is the AI for Good summit, organized by the United Nations’ International Telecommunication Union (ITU), wherever representatives from the backstage and nationalist sectors effort to sermon however to harness the exertion for the benefit, alternatively than the detriment, of humanity.

While Silicon Valley execs and AI laboratory leaders are testifying to lawmakers successful Washington astir the risks of superintelligence, and the White House slaps export controls connected chips, the UN AI for Good Summit—now successful its 10th year—is focused connected overmuch much idealistic goals.

“Our condemnation that artificial intelligence, deployed responsibly, could assistance lick humanity's astir pressing problems—from hunger to illness to a warming planet,” Doreen Bogdan-Martin, secretary-general of the ITU, said successful a keynote connected the conference’s main stage. “Today, that thought is being tested, including by the challenges AI itself is bringing, adjacent arsenic we strive to usage it for good.”

What bully means—and what bully it does humanity—was a question riven passim the conference, which dispersed crossed a monolithic 106,000-square-meter normal halfway connected the fringes of Geneva’s airdrome district. Sessions were backed by a drumbeat of interest that indifferent deployment by unchecked firm monopolies is already hardwiring planetary inequality and eroding quality rights.

For immoderate connected the beforehand lines, the tech industry's utopian veneer has already worn off. Speaking connected the sidelines of the event, Giulio Coppi, elder humanitarian serviceman astatine run radical Access Now, called retired the humanitarian and nationalist sectors’ overreliance connected large tech. “We should beryllium retired of the property of innocence,” Coppi says, demanding that organizations halt treating tech companies “as your champion friends.” He points to a decennary of opaque, multimillion-dollar deals funded by nationalist money. “You can’t adjacent explicate what's wrong your tech stack, due to the fact that it has kept changing,” helium warns.

Coppi’s absorption was muted compared to some: Pro-Palestine activists stormed the signifier during a keynote by Amazon main exertion serviceman Werner Vogels, alleging that the company’s exertion is being utilized by Israel against Palestinians, earlier yet being bundled retired of the venue.

“When we’re talking astir AI, we emotion the hype, we get excited astir it,” says Vijay Janapa Reddi, an engineering prof astatine Harvard University, implicit the din of competing sessions during a presentation. “The damn happening ne'er really lands successful practice.” The problem, helium says, is that “good” is excessively vague a modular to technologist against. “When you’re an engineer, bully means nothing. I can’t physique you thing that is good. A level that flies for 5 minutes ain’t nary good.”

Much of the planetary statement astir AI is present framed astir access: Who tin usage the models, who tin bargain the chips, and who is excluded from the compute economy. It’s portion of the crushed wherefore the Trump medication has implemented, past lifted, export controls connected starring frontier AI models, and China is reportedly mulling making its open-weight models little open. Tightening entree and cutting retired poorer countries tin permission them babelike connected overseas infrastructure platforms and standards.

In a league connected AI hardware and the widening integer divide, speakers argued that compute is nary longer simply a exertion problem, but a improvement problem. “If we mean AI for good, meaning compute for all, we indispensable admit that this is [about] improvement infrastructure, not conscionable technology,” says Syed Munir Khasru, president of the Institute for Policy, Advocacy, and Governance. Others pointed retired astir ample connection models stay structured astir English, making smaller, section LLMs moving connected cheaper hardware indispensable if AI is to service communities beyond the richest markets.