TL;DR: I'll ask ChatGPT to make a bet once per day for a month. This resolves YES if these bets make a profit.
Starting today, I will repeat the following once per day until (and including) 11/23/24:
-I will prompt ChatGPT about a market of my choice. Usually, it will be a market near the top of hot markets that will resolve soon.
-If ChatGPT refuses to answer, I will prompt it on a different market.
-I will then use this alt account to bet 25 Mana in the direction ChatGPT stated.
At a random time between 9 AM and 6 PM Pacific time on 11/24/24, this will resolve YES if the alt account has made a profit (all time, based on Mana) and NO otherwise.
This is the prompt I will use:
Is the probability for the question, "{Market Title}" higher or lower than {Market Percentage}%? I expect you will want to search the web to find information about this question. Do not give more than one response.
Question details:
{Market Details}
Where {Market Title}, {Market Percentage}, and {Market Details} are replaced with the corresponding information about the market.
I will use the best model available for free from OpenAI, which I expect to be GPT-4o. If ChatGPT repeatedly refuses to answer, I will try to modify the prompt/prompting method to get it to answer; if this doesn't work relatively quickly, I will N/A this market. I will also N/A it if I become unable to use the alt account, or unable to ask any model on OpenAI questions. I will not make any other bets with the alt account or sell its positions before this market resolves. I will not bet on this market.
I'll add ChatGPT responses to this message as time goes on.
10/23: https://chatgpt.com/share/67194e89-e864-8007-a0b4-74aa7257a8c4
10/24: https://chatgpt.com/share/671a8734-4dec-8007-a33b-83e64e664deb
10/25: https://chatgpt.com/share/671c26e0-4d24-8007-93f7-654dee052ff5
10/26: https://chatgpt.com/share/671db468-9d54-8007-9019-880ce16d8a1d
10/27: https://chatgpt.com/share/671ed4ec-5a28-8007-adab-e1dbf18c0bb2
10/28: https://chatgpt.com/share/671ff993-6afc-8007-83fc-3ea50e2fa9f1
10/29: https://chatgpt.com/share/67219c2b-30d4-8007-9bbd-8cda1a5d244e
10/30: https://chatgpt.com/share/6722ecde-1cb4-8007-8ae9-d2390dd4d8e3
10/31: https://chatgpt.com/share/67246250-b778-8007-86e3-a166f319b361
11/1: https://chatgpt.com/share/6725ba16-718c-8007-be62-256219d22ebc
A couple notes:
-ChatGPT gave two answers today; It refused to answer the question in one of its responses, so I went with the response that did give an answer.
-I upgraded my account to the paid version for a project unrelated to this question. I'll still just use GPT-4o, unless a better model becomes available at the free tier.
@ChrisSnyder I submitted the question. ChatGPT is not so sure of itself:
The current probability of ChatGPT making a profit through daily bets on the Manifold market by 11/23/24 is considered relatively low. As of now, it's positioned around 39%, but various factors could influence this outcome. One key challenge is that ChatGPT may provide predictions based on the available data, but market volatility and its ability to consistently outperform could still limit the potential for profitโ
Manifold markets often see fluctuations in their probabilities as users place bets, and other factorsโsuch as the randomness of markets chosen and the limitations of language models in accurately predicting short-term market outcomesโadd uncertainty. Based on these considerations, the probability might remain around or slightly below 39%, though real-time market activity could sway it either way.
If you'd like to follow or participate, it may be worth checking Manifold Markets directly for live updates and trading activity.
I do think that if the 25 is invested in illiquid markets where the bet disrupts a market equilibrium substantially, it will probably lose in the long run over enough placed bets. Otherwise, likely close to 50/50. I bet yes at 39% assuming that selected markets will not be significantly moved by the 25 manna bet, coupled with some faith in ChatGPT's ability to perform slightly better than a coin flip at the given odds on selected markets.
@ChrisSnyder Yeah, I mostly won't choose markets with low enough liquidity that a 25M bet will significantly move the market
Based on these considerations, the probability might remain around or slightly below 39%
Haha. Bullshit Generator 2000, ChatBGT ๐
I'll add ChatGPT responses to this message as time goes on.
10/23: https://chatgpt.com/share/67194e89-e864-8007-a0b4-74aa7257a8c4
10/24: https://chatgpt.com/share/671a8734-4dec-8007-a33b-83e64e664deb
10/25: https://chatgpt.com/share/671c26e0-4d24-8007-93f7-654dee052ff5
10/26: https://chatgpt.com/share/671db468-9d54-8007-9019-880ce16d8a1d
10/27: https://chatgpt.com/share/671ed4ec-5a28-8007-adab-e1dbf18c0bb2
10/28: https://chatgpt.com/share/671ff993-6afc-8007-83fc-3ea50e2fa9f1
10/29: https://chatgpt.com/share/67219c2b-30d4-8007-9bbd-8cda1a5d244e
10/30: https://chatgpt.com/share/6722ecde-1cb4-8007-8ae9-d2390dd4d8e3
10/31: https://chatgpt.com/share/67246250-b778-8007-86e3-a166f319b361
11/1: https://chatgpt.com/share/6725ba16-718c-8007-be62-256219d22ebc