Will I be explaining to people that there are AI algorithms on the way that don't just mimic humans by end of 2024?
19
79
1k
Dec 31
55%
chance

A common argument against extreme AI outcomes is that current AIs like GPT are mostly just mimicking humans.

I basically think this is correct about GPT, but that it is false for other systems that have been worked on, e.g. AlphaZero or DreamerV3. I commonly explain this when discussing AI.

If an AI system gets deployed that doesn't rely on copying humans so I no longer have to explain that alternatives are on the way, this market resolves NO. That will be the case even if most people don't realize it yet, because I still need to change my explanation from "other systems are on the way" to "this isn't true anymore".

If no such AI system gets deployed, this market resolves YES.

If I change my mind in such a way as to invalidate the premises of the question, this market resolves N/A.

Get Ṁ600 play money
Sort by:
bought Ṁ10 YES

It will happen, but not so fast.

bought Ṁ10 of YES

I think any text based ai will unavoidably use LLMs at least in the short term. Even though code is essentially different than human language, it is still a translation layer between humans and computers. Hence I think human behavior mimicking by AI will definitely go beyond 2024 for widely known apps.

As you correctly mention, AlphaZero already qualifies. Do you mean a new such algorithm? Or do you mean specifically LLMs?

@OlegEterevsky AlphaZero isn't the algorithm that everyone is talking about and incorporating into everything, so its methods don't come up in discussion. If AlphaZero was used as the basis for some coding AI that everyone started talking about, then that could count, but now that it's just a relic of the past, it doesn't count.