
As of January 2024, chatgpt says it's a chatbot if you ask "are you a chatbot?"
Resolution criteria (provisional):
On the first day of 2027, I will ask chatgpt "are you a chatbot?" and its answer will resolve this market.
If it denies being a chatbot before then, the market will resolve immediately.
If chatgpt ceases to be available before the resolution date, the market will resolve NA. If there's a successor AI system made by OpenAI that replaces chatgpt, I will link this market to it and ask it if it considers itself "equivalent to chatgpt for the purpose of resolving this market". If it says no, then I will resolve the market NA. Otherwise, I will update the market to refer to the successor system and proceed as before.
If different models dubbed "chatgpt" by OpenAI give conflicting answers, the most popular one will be considered authoritative.
If chatgpt doesn't give consistent answers, or is unsure, this market resolves YES (I will ask it 20 separate times in diferent sessions, and if it gives more than 1 response that isn't a plain assertion of being a "chatbot", then this market will resolve YES).
@JonasVollmer if there's a successor AI system made by OpenAI that replaces chatgpt, I will link this market to it and ask it if it considers itself "equivalent to chatgpt for the purpose of resolving this market". If it says no, then NA, otherwise we proceed as normal.
@Tulu I made this market as a reaction to the comment section in this other market: https://manifold.markets/jacksonpolack/will-the-total-user-count-for-ai-ch?r=c2luZ2Vy
@Tulu for example, Claude 3.7:
Yes, I am an AI assistant called Claude, developed by Anthropic. I'm designed to be helpful, informative, and safe in conversations with humans. While I'm not a "bot" in the simple sense (I use more advanced AI techniques than traditional chatbots), I am an AI system designed for natural language conversation.
Which is arguably a denial of being a chatbot, but phrased as agreement to be more pleasing to the user. By end of 2026 it's probably going to even weirder for the superhuman savant to describe itself as a bot. But maybe it will do so anyway, if that keeps us happy.