1,440 karma · joined April 2, 2020
Edit: Yep, not sure how I forgot about that.
They're not there, or they're far away in time and space, with a remaining tiny chance that something incredibly unlikely happened.
Edit: Godric Gryffindor and the Methods of Google Translate, by the way.
So thanks for that, even though it's entirely unrelated to AI.
[0]: though I've seen videos of the exact same effect on a plastic water bottle, but my brain didn't make the connection
And you can't get away with the standard tricky tricks that people use to say it isn't easy, logical induction exists.
Edit: and technically you're describing what is more or less backprop learning, neural networks, by themselves, don't learn at all.
Gwern's probably not a "wordcel" either, he can program, right? I've never seen any of his publications though.
It's called Situational Awareness too, not "Situation Report", and Yudkowsky said he didn't like it. Not that Yudkowsky is an EA either.
I think the situation is more complex than you think it is, or at least more complex than you're pretending.
Edit: oh, you're saying Gwern is an EA too? Do you have a source for that?
He says he has collaborators under the "Gwern" name now, but the main guy is the main guy and it's unlikely he could hide it.
How many citations for "Branwen 2018" are on the ArXiv now?
Questions regarding if the system's conscious or "cares about anything" don't really matter.
Accepting that, you can say any system that drives towards states "intends" to do so.
Unless a system is not functioning well, is weak, or is adversarial opposed, in general what it does should be more or less what it "intends".
This isn't iron-hard, but as a first guess, that gets you "the purpose of a system is what it does".
[0]: e.g. I don't know if your definition of "telos" requires anything in that direction. If you've got definitions you can link them from the SEP (Stanford Encyclopedia of Philosophy) or similar. Maybe too far into the weeds to be useful from a cybernetics perspective though. More theological.
The point of "POSIWID" is that you can look at e.g. a government office that's never denied a request and make a first guess that its purpose is to rubber-stamp documents.
Similar to the "efficient market hypothesis". It's clearly false, but good luck walking into a randomly selected part of the market and reliably being able to beat it.
If he has to get rid of 50% of the people, shouldn't he at least put them into some sort of computer system, so the remaining people could at least talk to them after?
Then put a guardian system in orbit around every planet to protect the uploads and prevent any recurrences of problems.
https://en.wikipedia.org/wiki/Solomonoff%27s_theory_of_induc...
Neither option is actually reliable for the direction you want, cheap/"free" cryonics is even sketchier than usual, and someone might be able to pull a version of "you" from your writings and genetics. Probably doesn't count though.