How about coming up with some policies with teeth that force the desired societal outcomes instead of pouring water on the fire before knowing what cooks? You can't slow R&D with policy, and model providers shouldn't be responsible for being daddy government and the police. When people do bad things with models, punish the people responsible. This is like wanting to restrict books because they have dangerous information within.
for what joy? and its posted on X , which is largely AI developed , ironical.. sorry . too late to that now .. we are well past that stage. and who will bell the cat .. everyone is an model developer, in some form or other.. and how much of ur post would reach this world fast filling with mass population working on ai projects ?
I largely agree with the concern, but "pausing AI" is an international effort requiring signature of treaties. Even then, enforcement mechanisms are lacking compared to, say, the number of nuclear bombers in a boneyard or the completion of nuclear tests.
From a moral standpoint is he wrong? Probably not. I think most of us can see that this technology has caused more harm than good. However this is another case of this man’s idealism getting in the way of objective realism.
Unilateral AI development pauses are incredibly dangerous. If one power pauses while others continue, it falls behind technologically and becomes vulnerable to adversaries with superior AI capabilities. This creates a prisoner’s dilemma where restraint punishes the restrained actor. Therefore, meaningful pauses only make sense with binding coordination across all major AI developers, which is a fruitless endeavour.
7 comments
[ 0.25 ms ] story [ 21.9 ms ] threadThis is silly.
How about coming up with some policies with teeth that force the desired societal outcomes instead of pouring water on the fire before knowing what cooks? You can't slow R&D with policy, and model providers shouldn't be responsible for being daddy government and the police. When people do bad things with models, punish the people responsible. This is like wanting to restrict books because they have dangerous information within.
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣀⣀⣀⣀⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀ ⠀⠀⠀⠀⠀⠀⢀⣠⣴⣾⣿⣿⡿⠿⠿⠿⢿⣿⣶⣦⣄⡀⠀⠀⠀⠀⠀⠀ ⠀⠀⠀⠀⣠⣴⣿⣿⣿⣿⣿⠏⢀⣤⣤⣤⣤⣿⣿⣿⡿⣿⣦⣄⠀⠀⠀⠀ ⠀⠀⢀⣴⣿⣿⣿⣿⣿⣿⣿⣤⣼⣿⣿⣿⠟⠛⣿⣿⣷⡈⠻⣿⣦⡀⠀⠀ ⠀⢀⣾⡿⠿⠿⠿⠿⣿⣿⣿⣿⣿⣿⣿⣿⣄⣠⣾⣿⣿⣧⣀⣘⣿⣷⡀⠀ ⠀⣾⣿⠁⢰⣿⣷⢀⣿⣿⣿⣿⣿⣿⣿⣿⡿⠟⠛⠛⠛⠛⠛⠛⠛⣿⣷⠀ ⢰⣿⣯⣤⣤⣭⣥⣼⣿⣿⣿⣿⣿⣿⣿⣿⠀⠀⣀⣤⣤⣤⣤⣤⣤⣽⣿⡆ ⢸⣿⡟⢻⡟⠛⠛⠛⠛⠛⠛⠛⠛⠛⠛⠋⠀⠀⣿⣿⣿⣿⣿⡟⠛⢻⣿⡇ ⠸⣿⣧⣾⣧⣤⣤⣤⣤⣤⣤⣤⣤⣤⣤⣤⣤⣾⣿⣿⣿⣿⣿⣀⣀⣸⣿⠇ ⠀⢿⣿⣿⣿⣿⣿⡟⠛⠛⠛⣿⣿⡏⠀⢘⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡿⠀ ⠀⠈⢿⣿⡿⠿⣿⣇⠀⠀⠀⣿⣿⣿⣶⣾⣿⣿⠿⢿⣿⣿⣿⣿⣿⡿⠁⠀ ⠀⠀⠈⠻⣿⣦⣹⣿⣆⠀⠀⢿⣿⡛⠛⠛⠛⠋⣠⣿⣿⣿⣿⣿⠟⠁⠀⠀ ⠀⠀⠀⠀⠙⠻⣿⣿⣿⣦⡀⠘⣿⣇⠀⠀⠀⣴⣿⣿⣿⣿⠟⠋⠀⠀⠀⠀ ⠀⠀⠀⠀⠀⠀⠈⠙⠻⢿⣿⣷⣿⣿⣷⣶⣾⣿⡿⠟⠋⠁⠀⠀⠀⠀⠀⠀ ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠉⠉⠉⠉⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
Unilateral AI development pauses are incredibly dangerous. If one power pauses while others continue, it falls behind technologically and becomes vulnerable to adversaries with superior AI capabilities. This creates a prisoner’s dilemma where restraint punishes the restrained actor. Therefore, meaningful pauses only make sense with binding coordination across all major AI developers, which is a fruitless endeavour.