Inside Mozilla’s AI crusade

With help from Derek Robertson

Mark Surman is putting his money where his mouth is when it comes to AI.

As president of the Mozilla Foundation — the nonprofit organization behind the Firefox web browser and other consumer gadgets and digital services — Surman has funneled most of his team’s efforts over the last five years into urging policymakers worldwide to embrace so-called “trustworthy artificial intelligence,” or pledges to use the technology for good.

And now Mozilla, at Surman’s urging, is taking bets with its own venture funds, in-house tech incubator and other for-profit gambles to show the world there’s a business to be made with trustworthy AI. That includes putting things like human rights, data protection and transparency at the core of how these complex systems work — in comparison to how almost all AI companies (including the Big Tech players) currently operate.

“We’re shifting gears into a broader approach on AI… in general,” he told Digital Future Daily. “We now have the commitment from all of our boards to figure out how we take our values and AI and build them into our thinking around what we do in the market as a social enterprise.”

That includes a $35 million venture fund to take bets on fledgling AI firms that mirror Mozilla’s ethos on AI. It’s not a lot of money compared to the billions flowing into more mainstream AI startups, but the nonprofit’s goal isn’t to compete directly with OpenAI. Instead, the nonprofit is seeking so-called “responsible AI” firms that focus on accountability, trust and safety. Recent investments include SecureAILabs, a startup using AI to keep patient data safe.

Mozilla is interested “more in use cases that matter for humans, but that aren’t getting paid attention to,” Surman added. “So we invested in Lelapa, which is an AI startup for Africans by Africans.” The Mozilla chief says many parts of the world — particularly those with languages that have limited interaction with the vast libraries of data required to build legitimate AI systems for these communities — are underserved by the new generation of generative AI startups, leaving a wide gap for someone like the nonprofit to step into.

Next is more in-house research and development on what trustworthy AI tools should actually look like. Surman said that years of lobbying politicians to take on the concept had left a hole in the market for people to make these principles a reality. Given Mozilla’s track record with the likes of Firefox, whose open source ethos contrasts to that of its larger rival, Google’s Chrome, he believes the nonprofit is primed to show what is possible.

“We’re looking across the board, moving from just the advocacy piece to we grow ourselves into being a player in the tech side,” he added.

Surman is conscious that even his organization’s substantial resources pale in comparison to those available across Silicon Valley. Still, he is pitching Mozilla as a trusted player — based on years of its advocacy work attempting to quell the possible downsides to AI — to sit alongside the likes of ChatGPT or Google’s Bard.

The goal is to leverage Mozilla’s relationship with policymakers and its sizable consumer tech know-how to show that trustworthy AI is a workable business, and that cementing human risks and tech accountability into this emerging technology can be a profitable business.

“If AI is defining the next era of our digitized work, and we want our values to show up in that, we need both a market based strategy and a movement based strategy,” Surman said.

nist's ai framework on the hill

The National Institute For Science and Technology is again a focus of attention in Washington, D.C., this time over its role with artificial intelligence.

In today’s edition of POLITICO Morning Tech for Pro subscribers, our team covered an oversight hearing where chip manufacturing and AI were in the spotlight. On the latter, members of Congress including Rep. Ted Lieu (D-Calif.) grilled NIST Director Laurie E. Locascio about what the institute can do to urge other parts of the Biden administration to adopt its AI framework released in January instead of developing their own piecemeal, something Locascio said she “couldn’t speak for.”

What she did speak on, however, was the institute’s efforts to work with the European Union in “harmoniz[ing]” AI rules transatlantically, something she said they were “actively engaged” in, and that she wants “to make sure that this becomes the way that people want to look at AI — responsible AI — globally.” — Derek Robertson

ai-powered soccer bots

The next job AI could be coming for: Professional soccer players.

Okay, well… maybe not professional. The bipedal robots that a group of DeepMind researchers trained to score goals, as detailed in a recent preprint — seriously — are unlikely to be out-dribbling your favorite Premier League striker anytime soon. But what they can do is seriously impressive. Just watch the videos of these pint-sized humanoids tottering around and kicking the ball through a net with eerie precision.

Jack Clark, the author of the Import AI newsletter, wrote on Monday that the quirky experiment “has some promising signs of life for transfer of complicated behavior from a simulated universe into our own real world.” Cameras help the robots orient themselves, and they’re programmed to teach themselves the most efficient way to score and learn new skills.

As the researchers write in the conclusion of the paper: “Even though the learned policies could be improved in terms of stability and perception, our results are encouraging, and we believe that similar methods could be applied to larger robots to solve practical real-world tasks.” — Derek Robertson

Tweet of the Day

the future in 5 links

Stay in touch with the whole team: Ben Schreckinger ([email protected]); Derek Robertson ([email protected]); Mohar Chatterjee ([email protected]); Steve Heuser ([email protected]); and Benton Ives ([email protected]). Follow us @DigitalFuture on Twitter.

If you’ve had this newsletter forwarded to you, you can sign up and read our mission statement at the links provided.