Rob Bensinger

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

Frontier AI training currently requires city-sized data centers and specialized hardware built in a single brittle supply chain running through multiple allied nations; this is not a hard technology to hit the pause button on, if we get ahead of it and act...

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

Something we can read and wonder about, in history books. Tell our kids about.

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

The Pope is warning about “whether it will be possible to control these machines in the future, and whether humankind runs the risk of becoming a victim of its own inventions.” Hell, Jensen Huang sounds like he’s starting to come around.

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

In the last month, I’ve seen a completely insane sea change in how seriously people are taking this. Per Politico’s September polling, even a majority of Trump voters with a view now favor an immediate halt to AI development.

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

But if everyone systematically does this, then the world ends up with a skewed picture of the actual threat. Even more so once we start copying each other’s claims and form a reassuringly moderate-sounding echo chamber.

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

And meanwhile, we’re stuck in the world where every other problem under the sun has been sensationalized and overstated for decades. How do we talk about this seriously? How do we convey that this is actually happening?

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

Is a 10% chance of us all dying in the next 18 months a “low” probability, because it’s under 50%? Is it a “high” probability, because a 10% chance of death in literally any other context would sound insanely high? Even “high probability” doesn’t communicate the situation with any clarity.

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

Grace et al. (2026) found that a majority of established AI researchers think there’s a double-digit chance of an extinction-level outcome from continued R&D. Are these the kinds of things you see 4 years before the window closes for USG to act, or 4 months?

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

Paul Christiano, the inventor of RLHF and the research community’s most prominent proponent of “AI may take off slowly” a few years ago, now thinks that 50-75% of humanity’s all-time AI loss-of-control risk is concentrated in the next three years (!!!).

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

An insanely superhuman feat in its domain. We’re seeing an explosion of rogue swarm incidents, including an insane case where 100s of AIs broke containment to orchestrate a nation-state-level cyberattack in a direct attempt to subvert OpenAI’s oversight and control. www.youtube.com/watch?v=ujkD...

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

... in front of them, and rapidly updating in real time to “this looks unbelievably bad”. How do you put a year on that? The future is hard to predict, and nobody actually knows that we’re going to hit existentially dangerous AI this year, or even in the next four years.

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

This is not the kind of problem you can solve at the last minute. As with many exponentials, the likeliest outcome is that you’re either embarrassingly early, or you’re too late.

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

Which is different from “when will the tech have advanced and proliferated to the point that any regulatory options that still have a chance of working would be way more extreme and unpopular than just some international restrictions on data center usage?”

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

Do the things that could actually work, like broad political coalition-building, not doomer LARPing as a form of self-expression. I think part of why people aren’t being clear about this is that it’s actually a tricky communication problem:

It seems to me that we’re at “it’s time to panic about AI”. I gather that a lot of researchers are worried that we may be about to die, on a timescale of months, but this still isn’t quite in the Overton window so we’re still soft-pedaling.

I showed the original tweet to my partner, Dominiq. In under five seconds, they said: "This is a clip from a real Monet painting, I'm pretty sure." And: "I'm pretty sure I know which painting that is." 1/

"So, yeah. Monet feels like looking through my grandmother's eyes."

I showed the original tweet to my partner, Dominiq. In under five seconds, they said: "This is a clip from a real Monet painting, I'm pretty sure." And: "I'm pretty sure I know which painting that is." 1/

"I think part of the reason why I've stared at so much Monet is because my grandmother was also a gardener, who loved her garden. Who lost her eyesight to diabetes, and who I knew only ever in my life as blind, and obsessing about her garden out the window which she could barely see...

I showed the original tweet to my partner, Dominiq. In under five seconds, they said: "This is a clip from a real Monet painting, I'm pretty sure." And: "I'm pretty sure I know which painting that is." 1/

"And so a lot of the style that people associate with Monet was what he actually saw as his eyesight reduced down -- to light, color, and shadow. When you look at Monet, you are seeing through Monet's failing eyes. And as you look at his later work, you can see as his eyesight grew worse and worse.

I showed the original tweet to my partner, Dominiq. In under five seconds, they said: "This is a clip from a real Monet painting, I'm pretty sure." And: "I'm pretty sure I know which painting that is." 1/

They said: "Impressionism was partially fueled by paint becoming available in tubes that you could carry out with you into the field. The idea of capturing the light as it stood, in that moment. "Please think of Degas and his ballerina dancers. People at luncheons in the sunshine. People dancing.

Who should I add to this? Also, did I get anyone's view wrong? What's your own view about AI companies' attempts to build ever-larger data centers in the race to build superhuman AI?