’They could use this to potentially make more scientific and engineering breakthroughs in a few months as the rest of the world would make in decades, and use these breakthroughs to supercharge their economy. Or, if they wanted to, they could do as much bioweapons research in a few weeks as it would take millions of the world’s top bioscientists to do in decades. ’
I think this is plausibly misleading though maybe not outright false. Certain kinds of research, for sure, involve only sitting around and thinking (for example, pure maths, but probably not only). Though in those areas you might be able to dramatically accelerate in speed, yes. But for a lot of research, actually doing experiments in the real physical world is probably necessary. For those sorts of research, just having AI will only speed up the bits that don’t involve doing experiments, which is going to cap the speed up at whatever amount of time doing the experiments actually takes (i.e. for example, if they take up 20% of the time, you can at most go five times faster.)
Though in fairness, it only takes 1 important area where doing things in the real world is not a significant time bottleneck for something dramatic to happen.
’They could use this to potentially make more scientific and engineering breakthroughs in a few months as the rest of the world would make in decades, and use these breakthroughs to supercharge their economy. Or, if they wanted to, they could do as much bioweapons research in a few weeks as it would take millions of the world’s top bioscientists to do in decades. ’
I think this is plausibly misleading though maybe not outright false. Certain kinds of research, for sure, involve only sitting around and thinking (for example, pure maths, but probably not only). Though in those areas you might be able to dramatically accelerate in speed, yes. But for a lot of research, actually doing experiments in the real physical world is probably necessary. For those sorts of research, just having AI will only speed up the bits that don’t involve doing experiments, which is going to cap the speed up at whatever amount of time doing the experiments actually takes (i.e. for example, if they take up 20% of the time, you can at most go five times faster.)
Though in fairness, it only takes 1 important area where doing things in the real world is not a significant time bottleneck for something dramatic to happen.
Yeah, you’re right, thanks for pointing this out. I’ll edit the post to reflect this.