On Wednesday, Google DeepMind did something a little unusual — it didn't launch a model, a chip or an app. It launched a place to argue. The DeepMind Institute is now live: a publishing platform for research and debate around artificial general intelligence, what AGI will do to jobs, safety and society. And the timing is no accident. Every week now brings a louder argument about whether AGI is already here, whether labs should slow down, and who gets to decide. Google, it seems, wants to host that argument on its own turf.
01.So what exactly is the DeepMind Institute?
Think of it as a serious, semi-academic blog with a heavyweight editorial team. The institute has three directors: Shane Legg, DeepMind's co-founder and Chief AGI Scientist, who also serves as managing editor; Demis Hassabis, the lab's co-founder and chair (and Alphabet's chief scientist); and James Manyika, Google's president of research, labs, technology and society. Researchers from Google, DeepMind and the wider research community will publish essays there — and the launch essay is candid that they "will not always agree."
There's a disclaimer too: nothing on the platform should be read as Google's official view. That gives Google cover to air uncomfortable ideas without committing to them. Smart, honestly.
02.The launch came with five essays
Day one shipped with a mission statement plus four pieces: one on reasoning transparency by Rohin Shah and Anca Dragan (how much of an AI's chain of thought we should be able to inspect), one on economic policy for AGI by Julian Jacobs and Alex Imas, a philosophy piece called "Principles for a new utopianism" by Stephen Cave, and a reprint of Hassabis's July framework for frontier AI — the same proposal for a FINRA-style standards body that fed into the safety coordination story we covered in detail earlier, where OpenAI, Anthropic and Google started exploring a shared safety standards body.
03."AGI hasn't arrived yet" — Legg pushes back
Here's the part that caught my attention. Nvidia's Jensen Huang said earlier this month that AGI has arrived. OpenAI made its own AGI claim on September 4. Legg, speaking to the Financial Times, isn't having it — he says it's premature to declare AGI achieved, and he's sticking with his long-standing forecast of a 50% chance of "minimal" AGI by 2028. What's striking is how stable that number has stayed through every model launch of the past year. The man who literally wrote his doctoral thesis on defining machine intelligence isn't budging.
04.The economics essay everyone's talking about
The meatiest launch piece is "Economic policy for AGI" by Jacobs and Imas, who rated 11 policy options for a world where AI displaces work. Their take: Universal Basic Income is "an expensive and blunt instrument" that may fail to concentrate relief where it's needed most. Instead, they like Universal Basic Capital — giving citizens a direct stake in productive assets, deployed as a backstop if labour's share of GDP falls for a sustained period. It scored highest of all 11 policies on personal agency (76.3 out of 100) but near the bottom on feasibility (33.1). In their survey of Americans, 85% backed publicly funded retraining while only 54% supported universal basic capital.
For a country like India, this debate isn't academic — it's exactly the jobs-vs-automation question our own policymakers are wrestling with right now.
05.A careful nod to the slowdown crowd
Legg also weighed in on Dario Amodei's recent call to slow — but not pause — frontier model releases, calling it "interesting directionally" and "worth considering". Read that carefully: it's warmer than a brush-off and colder than an endorsement. Notably cooler, in fact, than Hassabis's own "the direction is correct" on the same essay. It's a good companion read to our piece on Zuckerberg rejecting the slowdown — the industry is genuinely split on this one. Manyika, for his part, stressed that AI safety "can't be solved on a country-by-country basis," because the technology is global by nature.
Key Takeaways
- An AI safety debate platform called the DeepMind Institute is now live, led by Shane Legg, Demis Hassabis and James Manyika
- Legg stils it's premature to say AGI has arrived; he still puts 50% odds of minimal AGI by 2028
- One launch essay argues Universal Basic Capital beats Universal Basic Income for an AGI economy
- Legg called Anthropic CEO Dario Amodei's slowdown push "worth considering" — warmer than a brush-off, colder than an endorsement
Frequently Asked Questions
What is the DeepMind Institute?
Google DeepMind launched the DeepMind Institute on Wednesday — an essay platform for debating AGI, led by Shane Legg, Demis Hassabis and James Manyika.
When does DeepMind expect AGI?
Legg says it is premature to claim AGI has arrived, and he still puts a 50% chance of “minimal” AGI by 2028. One launch essay argues Universal Basic Capital beats Universal Basic Income for an AGI economy.
What This Means for You
Why should an Indian student or professional care about a Google essay platform? Because the questions it's hosting — will my job survive, what skills stay valuable, how does an AGI economy share its gains — are the exact questions that will shape your next decade. The institute's own framing assumes big labour-market dirruption is coming; the real question is policy, not whether. And when the world's top AI lab says AGI is "a few short years away" but its own chief AGI scientist won't shorten his 2028 timeline, that's your cue to stop panicking and start preparing. The practical move isn't waiting for universal basic capital — it's building skills that make you the person who directs AI rather than the person it replaces. Our GPT-6 Astra — Zero to Hero Guide is built for exactly that: prompting, agents, cost control and safety, taught from zero.