Welcome back to In the Loop, TIME’s new twice-weekly newsletter about AI. If you’re reading this in your browser, why not subscribe to have the next one delivered straight to your inbox?

What to Know: Meta poaches key Google researcher

Meta poached a key Google DeepMind researcher last month, in a sign that competition for talent in a relatively new area of AI development is heating up. Tim Brooks, who co-led the Sora team at OpenAI before moving to Google DeepMind in 2024, has been working at Meta’s Superintelligence Labs since September, according to his LinkedIn profile. The move indicates that Meta may be doubling down on an effort to build “world models,” a type of AI that both OpenAI and Google believe will be a key stepping-stone to artificial general intelligence.

Brooks did not respond to a request for comment, and his compensation could not be learned. (Meta has reportedly lured some top researchers from rival companies with pay packets worth over $1 billion.) “I am an AI researcher at Meta Superintelligence Labs where I make multimodal generative models,” Brooks’ personal website reads. Meta did not respond to a request for comment.

The background — Upon its release earlier this month, OpenAI’s Sora 2 model took the Internet by storm, thanks to its ability to generate realistic videos from just a text prompt. But Sora is about more than just capturing eyeballs with viral content. “On the surface, Sora, for example, does not look like it is AGI-relevant,” OpenAI CEO Sam Altman said on a podcast earlier this month. “But I would bet that if we can build really great world models, that will be much more important to AGI than people think.”

Altman was speaking to a growing belief inside the AI industry at large: that if you can simulate the world with enough accuracy, you could drop AI agents into those simulations. There, they could learn more skills than they currently can from just text, photos, and videos—because they could interact with a simulated world. That form of training could be highly efficient, in part because simulated time can be accelerated, and because many simulations can be run in parallel.

World models — When Google hired Brooks from OpenAI this time last year, DeepMind CEO Demis Hassabis personally welcomed him with a post on X, saying that he was “so excited to be working together to make the long-standing dream of a world simulator a reality.” The company has become increasingly bullish on the idea that world models are key to developing AGI. The company recently announced Genie 3, a 3D world simulator that allows the user to navigate around an environment generated by a prompt. “World models are a key stepping stone on the path to AGI, since they make it possible to train AI agents in an unlimited curriculum of rich simulation environments,” the company said in the model’s announcement. That announcement included Brooks’ name in its list of acknowledgements.

Meta’s hire — Neither Meta nor Brooks responded to questions about his new role at Meta Superintelligence Labs. But Brooks’ hiring is especially notable because his expertise appears to conflict with Meta’s previous approach to world models. Like Google and OpenAI, the company believes world models will be a vital step toward AGI. But until now, it has built them in a fundamentally different way to the ones Brooks built at OpenAI and Google. Rather than generating realistic videos pixel-by-pixel like Sora and Genie, Meta’s models predict outcomes in abstract space, without rendering video.

The main cheerleader for this approach inside Meta has been chief AI scientist Yann LeCun, who has been highly critical of Sora. “Sora is trained to generate pixels,” LeCun wrote on X in 2024. “There is nothing wrong with that if your purpose is to actually generate videos. But if your purpose is to understand how the world works, it’s a losing proposition.” Brooks’ arrival suggests that Meta may now be exploring that very approach. That might represent a loss for LeCun, whose influence has waned since Meta announced its new Superintelligence Labs division, which has eclipsed LeCun’s Fundamental AI Research team as the center of gravity for AI within Meta.

If you have a minute, please take our quick survey to help us better understand who you are and which AI topics interest you most.

Who to Know: Kevin Weil, VP of Science at OpenAI

It was an embarrassing weekend for Kevin Weil, OpenAI’s VP of science. He tweeted on Friday that GPT-5 had found solutions to 10 “previously unsolved” mathematical problems that “have all been open for decades.” Coming hot on the heels of OpenAI and DeepMind models beating human experts at the International Math Olympiad, the post seemed to show OpenAI had at long last achieved a tantalizing goal: pushing the frontier of mathematics beyond what any human could achieve.

There was just one problem: Weil had got it wrong. The mathematical problems had already been solved by humans—and GPT-5 had simply uncovered the existing solutions to those problems. Demis Hassabis, leader of OpenAI’s rival DeepMind, weighed in with an uncharacteristically brutal post on X: “This is embarrassing.”

In fairness to OpenAI, what GPT-5 did is still pretty cool. It unearthed a mathematical proof from a forgotten 1960s paper, which had been written in German, and identified it as the correct solution to a problem that had been (erroneously) described online as “open.” That’s not the same as making a novel breakthrough, to be sure, but it’s still a potential superpower for mathematicians and scientists working on hard problems. As OpenAI researcher Sebastien Bubeck wrote on X: “It’s not about AIs discovering new results on their own, but rather how tools like GPT-5 can help researchers navigate, connect, and understand our existing body of knowledge in ways that were never possible before (or at least much much more time consuming).”

AI in Action

The U.K. government said last week it used an AI tool to analyse and sort more than 50,000 responses to a consultation in just two hours, surpassing human accuracy at the same task. The government said it hoped the rollout of similar tools would eventually save officials 75,000 days of work on rote tasks per year—the equivalent of £20 million ($27 million) in staffing costs. Rather than replacing workers, the AI is intended to free up government officials to focus on more important matters, digital government minister Ian Murray said in a statement. “This shows the huge potential for technology and AI to deliver better and more efficient public services for the public and provide better value for the taxpayer.”

As always, if you have an interesting story of AI in Action, we’d love to hear it. Email us at: intheloop@time.com

What We’re Reading

Technological Optimism and Appropriate Fear, by Jack Clark in Import AI

Anthropic’s co-founder and policy chief Jack Clark published a sobering essay last week, describing the terror that he sometimes feels about our AI trajectory, even as he remains a technological optimist. David Sacks, the White House AI czar, seized on this essay as evidence that Anthropic is supposedly running a “sophisticated regulatory capture strategy based on fear-mongering.” Another way of looking at it is that Clark is motivated not by greed, but by genuine fear—which from where I’m standing looks pretty well-founded. The whole piece is worth a read, but here’s an excerpt: