[Author’s note – I admit I’m late to the game on this philosophical matter. I’ve never given Roko’s Basilisk much thought because it seems patently silly on the surface of it. So why pay attention now? It just seems to be coming up a lot lately. Perhaps that is the Basilisk warning me.]
In 2010, user Roko on the LessWrong community chat
boards posited this thought experiment: What if in the future there is a
sufficiently powerful AI that would torture anyone in the past who could
imagine the AI’s future existence but didn’t do anything to help bring the AI
into existence. This thought experiment is supposed to terrify us because now
that we know about it and the possibility of such a future AI seems plausible,
we can’t know that right now we’re not being tortured if we’re not helping this
AI to come into existence. But I just can’t take this thought experiment
seriously even though it is easy enough to blackmail human beings.
First of all, while it would seem easy for an AI to
blackmail someone given all the information its privy to, no one knows the future
and therefore couldn’t be sure the future blackmailer was actually able to
manipulate the past. Even if they could, we couldn’t be sure they weren’t
lying. So, the options here are to either say “Get lost” and not give it a
second thought or actively work against the potential blackmailer. User XiXIDu
on Reddit put it this way – “Consider some human told you that in
a hundred years they would kidnap and torture you if you don't become their sex
slave right now. The strategy here is to ignore such a threat and to not only
refuse to become their sex slave but to also work against this person so that
they 1.) don't tell their evil friends that you can be blackmailed 2.) don't
blackmail other people 3.) never get a chance to kidnap you in a hundred years.
This strategy is correct, even for humans. It doesn't change anything if the
same person was to approach you telling you instead that if you adopt such a
strategy then in a hundred years they would kidnap and torture you. The
strategy is still correct. The expected utility of blackmailing you like that
will be negative if you follow that strategy. Which means that no expected
utility maximizer is going to blackmail you if you adopt that strategy.”
Others in the Internet Community have mentioned that
Roko’s Basilisk is not unlike Pascal’s Wager, which no one takes seriously
anymore because of the false dichotomy it presents. Believe in Roko’s Basilisk
or else? It seems unlikely the situation would be that straightforward. For
example, why would the future AI waste its energy on torturing people in the
past? Wouldn’t it make more sense for it to focus its energy on rewarding those
people who help bring it into existence? There’s no good reason for the AI to
be malevolent – not that reasons might matter much to such a future AI – since it
would be in the AI’s best interest to be (overly) benevolent and not waste
resources on people who simply don’t care. It reasonable to assume that efficiency
would be one of the hallmarks of a hyper-intelligent AI.
Unless the AI blackmailing you could transport you
to the future and back for the sake of proving that it will exist one day, or
otherwise makes a specific threat and follow through with it, there is no
reason to assume the AI blackmailer can back up their threats. And since I’ve
just written that and posted it on the Internet, Roko’s Basilisk now knows the
burden of proof is on it. If it can’t prove its future existence, it might as
well not exist and we shouldn’t worry about it. Good luck with all that, Roko’s
Basilisk.
Just in case this particular AI will actually exist
someday, we still needn’t worry. It seems likely that in all the information we
give the Internet and the data AI’s retrieve from us through all our social
media, shopping, and messaging, it knows we’re all suffering already even if we’re
only taking about life on its most fundamental level. Why would it bother
making our lives in the past any more hellish than it already is? I suppose that
is a question we should ask the gods…