@deepfates Why? (I'd guess misaligned AI takeover is a bit less likely than not; you seemingly strongly disagree?)

We could probably eliminate the vast majority of these issues with a sustained, decently directed, and collective effort roughly as large as WW2. But will we?

@on_r3fl3ction You have no idea how much I literally mean everybody

There will not be a "last summer". There will not be a "permanent underclass". There will not be "human extinction". There will not be "endless suffering".

We are going to make it. Not because it's easy, but because it's possible. Because we can. Because we care enough to try

Doomer agi poasting is dead. The time for bloomer poasting has come.

@deepfates we're going to have a post-scarcity Star Trek-esque world where the idea of anyone going without at least a moderately luxurious existence is seen as pointless

@deepfates "We" is only a small segment of the world that is hyper-aware of what is coming, and is entirely clustered on X

@deepfates 💯 we need fix our default negative bias in our brain. we are just scared of new things. if we are rational enough, we will be optimistic about the future.

@thkostolansky @deepfates I suspect that things like the op tweet in aggregate result in a significant reduction in how well humanity handles AI. People want a reason to cope and keep doing what they are doing.

@deepfates No matter what the actual outcome is I think this is the most pleasant philosophy to live life by.

@deepfates Plenty of groups in history cared enough to try and were defeated by stronger forces. Caring is not enough. Preparedness is what we need

@RyanPGreenblatt Yes

@deepfates yes we do care. this is the kind of messaging that needs more amplification but also we do need a well defined vision and roadmap to get there. this is what we will work on

@deepfates When I say "roughly as large as WW2" I mean "political will roughly this high", costs would be lower.

TBC, you could reduce risk a lot with less will, but nearly eliminating risk seems to require a pretty big effort. (And we might do this! I think this is totally possible.)

@deepfates

@thkostolansky @deepfates uhhhhh

I believe: The persona-selection alignment strategy is bad. Far too many people believe the LLM mimicry with real behavior similar to a humans. Far to many safety folk have even forgotten how to think outside of the character frame. Laser focusing on hyperstition risks makes no sense, I don't think LLMs are the characters they are playing.

@deepfates @_nwyin i dunno man “hopium therefore p(bloom)>30%” is one thing but “p(bloom)=100%” is the kind of thing that may lead people to make the exact careless decisions that lead to doom

@RyanPGreenblatt @deepfates i think hes doing a hyperstition

@RyanPGreenblatt @deepfates i think there is value in being optimistic about the future that is strictly not actionable. i would hope that most people don’t see op tweet and think “phew ok i can stop thinking about ai now”