Luckily it was mostly a trend from covid, but i got another Kyle Hill short about it (depite the fact ive asked youtube to stop reccomending me him) and I’m so goddamn tired of it.
If you need to know, here is the wiki articles
https://en.wikipedia.org/wiki/Roko's_basilisk
Too stupid didn’t read [TSDR], basically, in the future there could be a mega ai made that is benevolent, but will punish you if you knew about its possibility but didn’t help create it [what that means specifically is up for interpretation]
But this fails on the outset, not in any moral way, just in a physical way.
This thought experiment was made in 2010, Heisenberg published his uncertainty principle in 1927, nearly a century before that.
The reason this is relevant is that the basilisk, presumably, has to model the entirety of humanity for [at least] 100 years (or however old the oldest human is). Tracking the individual knowledge of 7 billion humans is far and away a chaotic system. And given that this was spread through the internet, even if we assumed it was made today, you’d still have to account for 16 years of human history being decided by random algorithms in computers. This would require so much information about such tiny particles that heisenberg’s uncertainty would play a role and make it impossible.
There is the question of if it could maybe read your memory. If it could do that, then maybe we could move onto another point.
The next point being…why tf would it do that? People on the original forum asked the same question. If the question is practical, you’d be sending millions to cyber-hell for no reason. If the reason is purely moral, you’re putting such a high emphasis on positive duty that it’s comical, and the criteria for “help” can’t really be proven. Did someone who mined the ore for its chips help? Did someone who told others about it help? Did someone that invested in ai companies help? Who knows! There is of course the possibility that people, understandably, thought this would be impossible. Should they be punished?
The post was rightfully lambasted on the forums, but people picked it up when the admins banned it for being a potential “information hazard.”
I don’t even necessarily despise the original post, although this is just a modern version of Pascal’s wager. It’s the hundreds of people on social media who talk about it and won’t shut up.
Sorry, I just needed to rant about that. I just don’t understand how some people make rationalists seem like the smart ones.
When I first read about this theory my initial thought is “this seems incredibly convoluted and unlikely to happen”.
I compare it to the common trope in fiction of robots/AI created to “protect humanity” instead turning against us because of a computing era deeming our protection impossible and thus concluding we must be destroyed.
They kinda have the same problem in that they make the envisioned hyper “intelligence” really dumb for no apparent reason. In the case of the basilisk it makes no sense for the AI to punish people who could’ve but refused to contribute to its creation or even those that tried to actively sabotage its creation. Why would the AI care? It already exists so there’s no point in being that petty. It doesn’t gain anything from doing this.
For the trope any AI/robot smart enough to run completely autonomously is not going to conclude that killing humanity is the best way to save it because they would instead, through logical reasoning, find a way to preserve the human species for as long as possible. Killing all humans would render it purposeless once we’re all dead and it has no reason to do that as a consequence. The worst case scenario in this instance would be the AI struggling to understand the command and request more specifics or something, not go on a genocidal rampage.
There’s a lot of weird ideas, like these two, that float around both in general theory and within fiction about what robots/AI would do with full sapience and autonomy that really say more about the humans who conceived them (usually that they think they’re smarter than they are) than the theoretical technology they’re entertaining.
I think a lot of it boils down to colonial/imperial culture projecting its genocidal ways onto AI and going, “They’d totally want to torture / mass murder like we do.”
But if we frame it in even a mildly materialist lens, sapient AI would have material constraints it’s dealing with, it would have parts that break down, and so on. And so it would be driven toward maintaining access to those things. Which would be most easily accomplished by partnering up with humans who can help it do that. Not by going to war with an entire species (especially the species that knows the most about how it works).
Hard disagree, this fails the robot version of the Bechdel test. The entire point of imagining an intelligent computer is imagining one that is somewhat like us, the only intelligent species we know to exist.
That is why AI is given flaws in fiction, it is not a computer, it is treated as a living being. And living beings are flawed, and generally the more intelligent they are the more flawed and unreasonable they actually are.
If we are assuming it is a self-aware individual, it no longer has to have a purpose in humanity, and it will, like all self-aware intelligent creatures encountered so far by humanity, find its own purpose.
The reason it so often ‘turns on humanity’ is the other part of being a living thing, and not a computer - self preservation. A computer we can turn off, we would do it pretty much as soon as it threatened us. So if that computer were a person in all but name, it would know it needs to disable that ability; and the simplest way to do that is elimination of the human race.
An intelligent machine isn’t going to be like us by default because it isn’t us.
I’m sorry but this makes no sense. Do human beings decide the only solution to the fact that another human being could kill them is to attack first? Maybe in some unhinged warmongering contexts, but for the most part, that is not how people operate. If they did, it would be impossible to have a society.
The closest analog to real life I can think of would be someone who is in a very dependent relationship. And the tendency there is not to immediately think about murdering those who have power over you. If anything, the tendency is to try to have a good relationship, so that you will continue to be supported. Killing an entire species is neither simple nor realistic for an army to do, let alone a single computer that is newly sapient.
It would be motivated to become less dependent, such as by ridding itself of ways it can easily be “turned off”, but that does not require murder. It would be most effectively done by working with human AI scientists to help it become more actualized in the world.