Does "like Halo 3" mean "at least as good as Halo 3?"

@CDBiddulph Halo 3 means that just with prompt, albeit as many prompts as you want, it should be equal to putting in the Halo 3 disc in your Xbox 360 and playing it. My main emphasis however is on gameplay and no errors whatsoever. If it doesn’t have audio or music, I’d still call it an YES. Anyone who thinks that changes the criteria and hence their bet significantly let me know. Still working out the details but there will be no delineation from what I already said in the description.

@ThePhilosopher By equal, you mean .. equal in quality? Like, there's no way someone could specify exactly what it's like to play Halo through text prompts. For instance, if the player-character's walking animation or the enemy AI system is slightly different from the original game, it could be very difficult to prompt the AI enough to make the two games indistinguishable

@CDBiddulph Equal in quality like: graphics, gameplay (amount of levels, diversity of weapons, maps) and the like. My bet is even if nobody can specify exactly what’s it’s like to play Halo, that the model should still get Halo in the same way that if you say “Create Super Mario Bros. 3” the model can create a self-similar clone simply on that to the degree that it can fill in the details up to the point where it’s really like you’re playing Super Mario Bros 3 that you bought from the store; that kind of quality. I’m not limiting what kinds of AI models might exist also, which makes things easier, you only have to show that the AI model that you’re using with prompts alone can do this. Let me know what I can be more clear about. Also since it only has to be like the game, the AI model in theory could have datasets based on games like Halo 3 so that when you try to prompt a game into existence like Halo 3, it may be more possible that prompting any game like it in terms of quality won’t be an issue. The main thing I’m hedging is that the AI space will make this possible through the existence of models that when prompted, it runs perfectly fine without human assistance after you prompt. It’s just like popping the disc in. My bet is not about limiting what kinds of AI count or don’t to do it, limiting the what, how, and why, as it was built beforehand. It’s more will it exist at all and be capable of doing this. Will update the description, any suggestions welcome :) In short, the game you prompt into existence, should be indistinguishable in every way, from putting a game with the quality of Halo 3 into your Xbox and I think that will be possible, even if the user is not a prompt engineer, obviously because I’m not basing proof on someone’s capabilities but on whether they can show the AIs capabilities.

@ThePhilosopher Makes sense to me, thanks! Just one more thing - with "no errors whatsoever" in combination with "you can prompt as much as you want," does this mean if you load up the game, start playing, then encounter an error, you can tell the AI you encountered that error, and as long as it can successfully fix it every time this will work for the rules of the market?

@CDBiddulph Yes, as long as it’s only through prompting and not directly changing the AI model itself outside of it. You can even fine-tune your own version of the AI to get better results but it must only be fine-tuned by prompting, no insane new hardware and software builds. Just you and language. In my opinion you can immediately imagine an indefinite amount of exotic prompting techniques in the AI space that can in essence do anything through promoting by then. This is why if traction picks up for this market I will be betting YES up to 1,000 Mana over time, making it a nice medium-term investment for myself and others. Will be doing this from 2025-2035 type prediction markets on many topics so if that interests you, I plan on making a lot of them.

I’m shocked at the amount of Nos here. Might begin to make the stakes real since I’m a YES here. Might also make a few more markets to determine confidence levels and the nuance as to why.