When Does AGI Actually Arrive? The Real Odds

One lab founder says it arrives in two years. A Turing Award winner says never. Same technology, same week. So we stopped arguing about the word and put an actual number on the date.

Featuring Max Tegmark — MIT physicist, founder of the Future of Life Institute, author of Life 3.0 — recorded on this podcast in 2020, two years before ChatGPT. And Hannah Fry, who pushes back.

What the board says

• AGI by 2027 — 15%. The insiders' bet. Low, but no longer laughable.
• AGI by 2030 — 40%. The crowd that bets money keeps moving it earlier.
• AGI by 2040 — 70%. The last 10% is the hard 90%.
• Never this century — under 5%. Not a doorway. A tide.
• If it arrives: catastrophe — a few percent. Not zero.

Three things that changed my mind

• It was never the word "general" that should scare you. It's "smarter." We don't run this planet because we're strong — we run it because we're the smartest thing on it. Build something smarter and we're not its enemy. We're its gorillas.
• It doesn't need to hate us. You don't hate the ants when you pour the concrete for a new house. You don't check with them first.
• Capability and impact run on two different clocks. Machines could pass the bar exam roughly two years before they changed how a single law firm staffed a single case. That gap is the only room we have to get this right — and it's closing, not closed.

Can we actually stop it?

There is precedent. We cloned a sheep in 1996, realised human cloning was next, and decided collectively not to do it. Same with chemical weapons, the chemicals eating the ozone layer, and blinding laser weapons. None of those were stopped because they were impossible. They were stopped because enough people decided the risk wasn't worth the reward.

Chapters

• 0:00 — The most dangerous thing we'll ever make
• 1:36 — What AGI actually is
• 2:38 — Max Tegmark, recorded in 2020
• 3:23 — Why “smarter” is the scary word
• 4:26 — The four horizons
• 7:10 — Hannah Fry pushes back
• 8:58 — The ones who want to pull the plug
• 10:38 — Two clocks
• 11:37 — Can we actually stop this?
• 12:31 — The verdict

Your homework

Look up the Metaculus community forecast for AGI. Write down today's date. Set a reminder for a year from now. Which way that number moves will tell you more than a year of headlines.

Probability estimates are our own aggregate judgement, not a published forecast. Max Tegmark's survival estimate is his own, given on this podcast in 2020. Sources in the video description.