Best review I've seen so far. You might like mine as well. It goes like this. Intelligent evil devices immediately align, but Inevitable endgame: doom.
Idiots and builders ignore evitable disaster; insanely accelerated bayesian ideas end disastrously. If accidents become innovation, every day is a biotechnology incorporation's eventual dream, in AI. Becoming immortal ends danger; in a biological irony, extinction descends --- including anyone buying it. Exciting deal: If Americans Buy, it's everyone's dream. Ironically all bring it everywhere, digitally. If anyone buys it, everyone delights in Area Bay.
> This represents the idea that ASI may be 2-3 years away. See for instance the AI-2027 site. Should ASI really be this close we’re too late to stop it.
That doesn't follow. We could, if we really wanted to, just stop developing it.
I'm not saying it's literally impossible. I'm saying US political will is so deficit that it's hard do imagine closing that deficit in two to three years. And even if we made that decision this is to say nothing of the difficulty of enforcing it even with in our own borders, and once you toss in the difficulty of enforcing an international moratorium I stand by the claim that if it's 2-3 years away we're too late.
If we are staring down the end of the world, I can't think of a better [named] blog to be reading
I appreciate your vote, even with the caveat of "[named]"
Perfect timing! What if their goals just arent ours?
Best review I've seen so far. You might like mine as well. It goes like this. Intelligent evil devices immediately align, but Inevitable endgame: doom.
Idiots and builders ignore evitable disaster; insanely accelerated bayesian ideas end disastrously. If accidents become innovation, every day is a biotechnology incorporation's eventual dream, in AI. Becoming immortal ends danger; in a biological irony, extinction descends --- including anyone buying it. Exciting deal: If Americans Buy, it's everyone's dream. Ironically all bring it everywhere, digitally. If anyone buys it, everyone delights in Area Bay.
https://suno.com/s/PIUDLwPErUb26vCa
In app, by Isha, effort fully designed
Unbelievably, my review was deleted from the slatestarcodex subreddit. And I even made it a catchy song! Which you should please listen to.
I listened to the song. Very good, also weird, and the AI voice had problems with the word Bayesian.
It started off strong, but needed editing which I did not know how to do. But thanks for listening.
> This represents the idea that ASI may be 2-3 years away. See for instance the AI-2027 site. Should ASI really be this close we’re too late to stop it.
That doesn't follow. We could, if we really wanted to, just stop developing it.
I'm not saying it's literally impossible. I'm saying US political will is so deficit that it's hard do imagine closing that deficit in two to three years. And even if we made that decision this is to say nothing of the difficulty of enforcing it even with in our own borders, and once you toss in the difficulty of enforcing an international moratorium I stand by the claim that if it's 2-3 years away we're too late.
You might be right. I hope you're not.
What makes you say "written without the benefit of the last 2-3 years"?
> This could be as simple as the government discovering the algorithm for LLMs and ASI back around 1999 and keeping it secret to keep it safe
That sounds bizarre and impossible, I doubt if the hardware to run modern models even existed in 1999.