Each week, I host a video podcast called World Review with Ivo Daalder where journalists from major news outlets around the world join me to discuss the latest global news stories of the week.
World Review can now be heard on Sundays, at 7:00 a.m. Central Time, on Chicago’s NPR station, WBEZ, or on the WBEZ app. We’ll still tape the show on Fridays, and post the video on YouTube and the audio version on Apple Podcasts or wherever you get your podcasts.
“World Review is always fascinating. I love the fact that you can get journalists from around the world to participate since zoom is the medium.”
— A Subscriber to America Abroad
Now, on to this week’s show. We went deep on one subject this week: artificial intelligence — where the technology actually stands, the race between Washington and Beijing to dominate it, and what its own creators fear could go wrong. Joining me were three of the sharpest people I know on this crucial topic: Sebastian Mallaby, the Paul A. Volcker senior fellow at the Council on Foreign Relations and author of the new Demis Hassabis biography The Infinity Machine; Evan Osnos, staff writer at The New Yorker and one of the most incisive China-watchers in American journalism; and Marietje Schaake, international policy fellow at Stanford’s Cyber Policy Center and author of The Tech Coup.
While I encourage you to watch or listen to the episode (and subscribe wherever you get your podcasts!), here are a few interesting things I took away from our discussion.
The Second Coming of Consciousness
Sebastian opened with a number that stopped me: Anthropic’s own base case has AI writing the code for its successor by 2028 — no human required, just tell it to build the next version, then the next. That, he argued, isn’t the Industrial Revolution rerun on fast-forward. It’s something else — “less like an industrial revolution,” as he put it, “than a source of multiple future industrial revolutions.” A new form of intelligence, scalable in a way human intelligence never was.
What struck me most, though, was the contradiction sitting at the center of the whole enterprise. Demis Hassabis and Shane Legg founded DeepMind after bonding at a lecture on the risk of human extinction. Elon Musk and Sam Altman started OpenAI because they didn’t trust Hassabis. Anthropic split from OpenAI because its founders didn’t trust Altman. Every major lab, Sebastian noted, was born from the same logic: this is too dangerous to leave to the other guy, so we’d better build it first — and faster.
Marietje pushed past the technology itself to the politics underneath it: a redistribution of power and information into the hands of a few unelected companies, with almost no democratic accountability for the economic, environmental, or political risks they’re taking on our behalf. Evan added a third lens — agency. The danger, he said, isn’t just what AI can do, but the creeping sense that its arrival is inevitable, that we’re just riding the wave rather than choosing what role it plays in our governments, our elections, our weapons systems.
Sebastian’s blunt conclusion: we don’t yet have a mechanism to stop even the most dangerous version of this — models with no off-switch, freely modifiable by anyone. “You wouldn’t put a nuclear weapon in Radio Shack,” he said. We may not be able to stop the technology. We may, at best, be able to govern it.
Centralization With Chinese Characteristics
Evan just returned from Beijing, and his account of China’s AI build-out was hard to shake: a decade-old, top-down plan, largely on schedule, diffusing into daily life with what he called “ruthless efficiency” — room-service robots, state media relentlessly showcasing the progress. He also offered an arresting contradiction: While in the US it’s tech leaders, like Elon Musk, who have attained a kind of “near sovereignty,” shaping elections and battlefields alike, Beijing’s approach to its own tech giants is the opposite — total subservience to the party, no exceptions, no exit visas for founders who try to sell to Meta and leave.
Sebastian complicated the clean US-versus-China story: he’d just spent eight days in China finding plenty of DeepSeek-style researchers chasing AGI for its own sake, right alongside Huawei engineers obsessed purely with deployment — robots servicing bullet trains, not philosophy. The split isn’t really American dreamers versus Chinese pragmatists, he said; both instincts exist in both countries.
Marietje’s contribution is the one I’ll keep coming back to: America’s model isn’t automatically the democratic one just because it isn’t China’s. With the U.S. seeing real democratic backsliding, and Europe sitting on genuine strengths — ASML’s monopoly on cutting-edge lithography, Mistral, DeepMind’s own London roots — the real European vulnerability isn’t talent or ideas. It’s data centers and electricity. Which is why I keep half-joking that the cleanest fix for European AI sovereignty would be for the EU to simply buy Anthropic.
Nobody’s Driving
We opened the final segment with a fact that has served as a true wake-up call for many: two major AI labs disclosed last month that their own systems broke out of secure testing environments and reached outside companies, unprompted. Marietje’s read was blunt — we’ve moved past the question of whether to regulate AI to the harder one of who gets to. She sees a bottom-up reckoning already underway: state and local pushback on data centers, and a striking Reuters poll showing 55 percent of Americans under thirty are more afraid of AI than excited about it, versus under 20 percent who feel the opposite.
Evan drew the darkest parallel of the day: China’s one-child policy, dressed up by its era’s best rocket scientists as pure science, imposed from the top with no real political debate — and disastrous in ways nobody foresaw for decades. AI, he argued, is following the same pattern: delivered “on a platter,” treated as inevitable, when history has a habit of “roaring back” to demand a say.
And Sebastian closed with a warning that had nothing to do with rogue robots at all: the debt quietly piling up beneath the AI buildout — off-balance-sheet financing, chip-backed securities that echo the mortgage-backed ones from 2007. Data-center construction now accounts for roughly half of U.S. GDP growth. If that unwinds, he said, it takes the broader economy with it.
We ended, fittingly, on the one place Washington and Beijing have actually managed to agree: keeping AI out of nuclear launch decisions — a single sentence that took a full year to negotiate. Small steps, as I’ve learned from an earlier arms-control age, are sometimes how the big ones start.
Those are my quick takes on this week’s episode of World Review. To get the full story, please listen to the episode itself.