No. Some other off-ramp will be needed.

AI research progress probably can’t be stopped completely. With enough time, researchers would probably eventually figure out much more efficient methods of creating AIs.^* Or perhaps, with enough time, some rogue actor would eventually succeed at subverting a ban.

Time will most likely drag humanity into the future one way or the other. And humanity will either go extinct — as most species have before it — or will somehow navigate the transition into a world where smarter things exist.

But humanity also doesn’t need to buy time forever. AI is not the only technology that’s progressing. Biotech is also starting to mature, and if humanity manages to prevent the development of superintelligent machines for multiple decades, it will have to contend with upsets such as genetic engineering that results in significantly smarter humans.

The question is how much time we can buy, and what we can do with that time.

The basic problem that humanity faces is that of safely crossing the gap from human intelligence to superintelligence. The best plan we can think of that sounds like it maybe has a chance of working in real life is to buy time for biotechnology to augment human intelligence quite a lot — to the point where future human researchers can get so smart that they’d never (for example) estimate an engineering project would finish on time and under budget unless it actually would.

So smart that they’d never commit to a scientific theory like Aristotelianism or heliocentrism, even if the society surrounding them was completely convinced. So smart that they’d have a chance at navigating the gap between Before and After on the very first try.

There are other possible paths forward we could imagine, but this one has the advantage of attacking the key bottleneck (“the existing scientific community is too reliant on trial-and-error methods and incrementalism to handle this particular problem”), using tech that is already beginning to be available today, without posing a serious risk to the world.

A monitoring regime shouldn’t last forever.

It’s theoretically possible for humanity to balance on the knife’s edge of competence, where we currently sit, forever. Our guess is that this would require draconian control of people’s thoughts and activities. But even if it didn’t, we would consider it a poor choice.

We, personally, think that humanity’s descendants deserve to become whatever they wish to be, explore the stars, and build a flourishing and beautiful civilization there. We advocate for a ban on frontier AI development because we think superintelligence is dangerous enough to make this necessary — not because we hate AI, or technology, or scientific progress.

The real question is how we get to a wonderful future, and how we manage the transition from here to there.

This is worth stressing in part because there are many people who present AI as a false dichotomy: They say (falsely) that society must either accept the risks of AI and plow ahead at full steam, or reject AI and let our civilization fade out on a single planet forever. This is simply wrong. There are other routes to the future — routes that permit a future that’s just as bright, but without nearly so high a risk of throwing it all away for nothing. Humanity should find some other path to the future.