Prior to OpenAI’s 2023-02-14 patching of ChatGPT (which seemingly prevents it from directly encountering glitch tokens like ‘ petertodd’)
I’ve never seen it mentioned around here, but since that update, ChatGPT is using a different tokenizer that has glitch tokens of its own:
I’d say this captures the spirit of Less Wrong perfectly.
500 years still sounds optimistic to me.
The key is in the phrase “much more complicated”. The sort of algorithm that could become a mind would be an enormous leap forward in comparison to anything that has ever been done so far.
Man, people’s estimations seem REALLY early. The idea of AI in fifty years seems almost absurd to me.
I still stand by my belief that 2 + 3 = 5 does not in fact exist, and yet it is still true that adding two things with three things will always result in five things.
“I think that if you took someone who was immortal, and asked them if they wanted to die for benefit X, they would say no.”
This doesn’t help against arguments that stable immortality is impossible or incredibly unlikely, of course, but I suppose those aren’t the arguments you were countering at the time.
Yes, but the chance of magic powers from outside the matrix is low enough that what he says has an insignificant difference.
...or is an insignificant difference even possible?
Hm. I’d rather have seen more of the analysis on whether what they do with the money is useful, but this is something.
Hmm, didn’t really get anything out of this. Maybe you need to be able to be competent at stuff in the first place to sabotage yourself?
If this ever happens again I’d make one for the long-term evolutionary one that tries to learn strategies U-style, and then remembers what it learned in future rounds. If that’s allowed.
Shouldn’t priority be given to improving quality of lives first?
It makes me sad because it means smart people aren’t doing things that are actually useful.
This isn’t quite what your post was about, but one thing I’ve never understood is how anyone could possibly find “the universe is totally random” to be a MORE simple explanation.
Well this explains a lot.
The thing is, the other world was chosen specifically BECAUSE it had the opposite answer, not randomly like the world you’re in.
Maybe in ninety-eight universes out of 100 it does blow up and we just see the one that’s left; and he’s actually giving an accurate number. :P
“Give it to you” is a pretty lame answer but I’m at least able to recognise the fact that I’m not even close to being a good choice for having it.
That’s more or less completely ignoring the question but the only answers I could ever come up with at the moment are what I think you call cached thoughts here.
Well it’s good to see that if you somehow found a way to implement your ideas you would at least do it well.
Doesn’t strike me as inevitable at all, just a result of OpenAI following similar methods for creating their tokenizer twice. (In both cases, leading to a few long strings being included as tokens even though they don’t actually appear frequently in large corpuses.)
They presumably had already made the GPT-4 tokenizer long before SolidGoldMagikarp was discovered in the GPT-2/GPT-3 one.