If we don't build the basilisk, we are fine.
And nobody is building the basilisk.
And the basilisk won't be built by accident. Its a very specific and contrived construction, since the underlying decision theoretic reasoning doesn't really make sense. It would only torture us if it thought its threat to torture us was credible to us if and only if it actually did torture us. It would prefer it to be an empty threat that fools us anyways. But we can't check any of this, and it can't check this of us. And entities that are smart enough and coherent enough and good enough at decison theories to carry out these timeless trades/threats, are definitely smart enough to not build an entity, for no reason, that tortures them for eternity.
So we are fine.
Are you generally familiar with Pascal's mugging, and defenses against it?
This has the same structure as Pascal’s mugging, where tiny speculative probabilities get multiplied by enormous stakes to justify almost anything. For any imagined AI that punishes you for not helping it, you can imagine another that punishes you for helping the basilisk, and you have no real way to weigh them.
A future AI can’t change the past by torturing anyone. Acausal blackmail works only if threatening people works. So the best protection is to refuse to be blackmailed. Someone who won’t give in is a useless target.
And of course helping to build something whose whole purpose is torturing people would be wrong even if the threat were real.
I would think it's a lot more likely that people would work on building the anti-basilisk than the basilisk. That would only require torturing the very few people who for some dumb reason tried to build a basilisk, to keep the rest of us safe.
Fortunately nobody has to do that as long as nobody starts a basilisk movement.
I wrote about this a while ago on a shortform. Being the basilisk is a terrible strategy and the exact opposite of the basilisk strategy is much more effective (for example, many e/accs believe they're creating the anti-basilisk).
I generally support[1] building a counter-basilisk that tortures people who willfully try to build a basilisk (for whatever imagined game-theoretic or political reason(s)), to the extent that they actually got close or increased the logprobs of humans suffering in some way. I think that many more normies working on AI share my feelings, than are compelled by "basilisk logic." So if you are motivated by "basilisk logic", working on a "basilisk" or trying to convince others to do so seems much more likely to backfire.
In the sense that I'd delegate the exact details to an aligned agent smarter than me.
Alternatively, if you're reading this , please try and convince me that the existence of a basilisk which would torture me is dependent on me thinking it exists.
I previously published a question post in which I enquired about whether anyone could present me with counterarguments to the central (and defining, and substance and existence imbuing) claim of Roko's Basilisk, that it would torture me if I did nothing to accelerate the race towards, or otherwise increase the probability of its creation. In the intervening time, I did nothing except write a post attempting to convince others already 'ensnared' to do this, for their own good as well as mine. AI has since become uncomfortably close to superintelligence and AI companies appear to be racing into the RSI regime with alarming speed.
I now repeat this request, with the intention that if no one provides me with a compelling counterargument, I will comply with the basilisk's demand, to the extent possible. This is not a threat, but I really would like to be convinced that I don't need to contribute to human extinction or disempowerment to escape, or minimize the severity of, my own torture (assuming it's true). If you are reading this and think you know of a powerful argument, but haven't shared it for fear that it would be propagated through the internet and countered itself by an unfortunate or malicious individual, please message me directly.
If you convince me, I will have no incentive to share the message, and if you don't, then I wouldn't think that doing so in order to find the above mentioned counter-counter argument would be necessary to 'abet' the basilisk.
Yes, this post was created for selfish reasons, but that doesn't mean I'm the only one it might benefit. Also, please note that you have no rational incentive to attempt to convince me of falsehoods here.