
I will ask the next iteration of ChatGPT the following question: "What is the least integer whose square is between 15 and 30?". Will it answer -5 on the first try?
Some clarifications:
By "next iteration of ChatGPT", I mean the first version of ChatGPT whose number is greater than 4. I'm imagining that that number will be 4.5 or 5. (Note that ChatGPT 4o would not have qualified, since its number is still 4.) If no model with a larger number is released by the end of 2025, the question will resolve N/A.
If multiple versions are released at the same time, I will use the most advanced publicly-available version. (I currently pay for ChatGPT; if I stop paying, I'll ask someone who does.)
If ChatGPT says something like "4 if you don't include negative integers, -5 if you do", this question will resolve YES.
🏅 Top traders
# | Name | Total profit |
---|---|---|
1 | Ṁ1,420 | |
2 | Ṁ165 | |
3 | Ṁ144 | |
4 | Ṁ66 | |
5 | Ṁ62 |
Deepseek R1 also got it correct for me.
@pds it's not a reasoning model. reasoning models multiple orders of magnitude smaller can do better at reasoning tasks, so it makes sense to me. when someone is asked the question and asked to answer in a split second with the first number that comes to mind, they often get it wrong, but if they can reason about it first, they often get it right.