Here is what I would tell someone if they asked me what I learned from “The AI Doc: Or How I Became an Apocaloptimist,” which premiered at Sundance and opened in theaters on Friday:
There are now 20,000 people working on artificial general intelligence, and fewer than 200 working on alignment and safety.
Those numbers induce anxiety. Numbers, however, rarely tell the whole story.
In the 104-minute documentary, directors Daniel Roher, the 2023 Academy Award winner for Best Documentary for “Navalny,” and Charlie Tyrell, try to make sense of the rapid onset of AI.
It features several of the CEOs driving the AI boom, including OpenAI’s Sam Altman, Anthropic’s Dario Amodei, DeepMind’s Demis Hassabis, high-profile researchers like Jan Leike and Ilya Sutskever, and several experts on AI ethics and risk. Elon Musk, the filmmakers say, got too busy.
The premise is personal. Roher is preparing for the birth of his first child and, like many parents, wants to understand the kind of world into which he’s bringing a new life.
The result is illuminating but unsettling. There’s a torrent of competing motivations driving this latest wave of technology, and it’s unclear which will win out. Here are a few things I learned from the documentary.
Intelligence is risky
The film’s main characters make clear that the industry’s ambitions are huge, and that we’re only at the start of a seismic shift.
“This is just a warmup. The really powerful systems are still coming, and they’re going to be coming quite soon,” Shane Legg, the chief AGI scientist and cofounder of Google DeepMind, tells the filmmakers.
But the film doesn’t linger on ambition without immediately confronting its consequences. As researchers describe systems growing more powerful, the conversation turns to what happens if those systems aren’t aligned with human interests.
Connor Leahy, an AI researcher and founder of Conjecture, a UK-based AI alignment startup, said there may come a day when AI responds to humans the way humans respond to ants. “We don’t hate ants, but if we want to build a highway and there’s an anthill there, well, sucks for the ants,” he said.
The safest path forward for humans might be a future where humans and machines are intertwined.
“We’re going to merge with AI. We’re going to merge with technology,” Peter Diamandis, a longevity investor and founder of the XPRIZE Foundation, tells the filmmakers. “By the early to mid 2030s, expect that we’re able to connect our brain to the cloud, where I can start to expand access to memory.”
The AI race is all about profit
The AI race is driven by profit, and several of the film’s subjects said prioritizing safety — at least right now — only comes at the expense of the bottom line.
“If two people are in exactly the same place, the one willing to take more shortcuts on safety should get there first,” Altman, wearing a brown sweater and sitting against a charcoal black screen, tells the filmmakers. “But we’re able to use our lead to spend a lot more time doing safety testing.”
That’s a stark framing of the trade-offs at the heart of the industry: speed versus caution, competition versus collaboration. In a race to build the most powerful machines, the incentives reward those who move fastest, not those who move most carefully.
Journalist Karen Hao, author of “Empire of AI,” reinforced that idea, telling the filmmakers that “profit maximization incentives” are the primary force behind the AI race.
That’s all a little concerning.
You should still have children
Despite the fear around the arrival of superintelligent machines, however, Altman struck an optimistic note about the future of life itself.
“Having a kid is just this momentous thing that I stay up every night reading these books about how to raise a kid,” he said. “I hope I’m going to do a good job, and it feels very overwhelming, but I’m not scared for kids to grow up in a world with AI.”
The film wraps with a meta-montage of Roher and his wife debating how to end the movie, welcoming their newborn, and having emotional conversations with their parents — all of which feels a bit rushed, and scripted, almost like AI wrote the final scenes.
At the end, I turned to the elderly woman sitting next to me, who was fast asleep and gently snoring. It was like we were watching a weird bedtime story: The risks are real, even as a new and better world looms.
Read the full article here
