AI Emergency: The AI Labs Are Lying To Everyone, He Says 99% Chance Of Extinction | Roman Yampolskiy

A four-way debate, which the title does not say. Ed Zitron, Andrew McAfee, Nate Soares and Roman Yampolskiy each wrote a probability of extinction in an envelope; the headline takes the highest of the four and hangs it on one name. The spread is the substance — the host reads out 99% and a rounding-error zero within seconds of each other.

Yampolskiy’s argument is not that the odds are bad. It is that the problem is the wrong shape: even the safety community, he says, assumes that with more time and more money and smarter graduates, superintelligence could be controlled indefinitely — and his research points the other way. He compares it to building a perpetual motion device, a system that must never make a single mistake across every interaction, every malevolent actor, every round of self-improvement. If that is right, pacing does not help, and the conclusion is a permanent ban on general superintelligence while narrow systems continue. Nate Soares argues the same corner from MIRI; his book is called If Anyone Builds It, Everyone Dies.

The other half of the room rejects the frame rather than the numbers. McAfee puts himself at approximately zero, says he does not think LLMs are the path to superintelligence, and calls the extinction discussion a distraction from the substantive conversations about what AI is actually doing. Zitron presses the same objection from the harm side: oxygen spent on what might happen is oxygen not spent on people being manipulated and hurt now. The debate opens on the OpenAI–Hugging Face swarms, which all four accept as fact and none of them read the same way.

These are auto-captions, so nothing here is quoted verbatim.