Your experiences match mine perfectly. Probably the most disconcerting ones for me is its enthusiastic encouragement when I've gone off track and the "miracle occurs here" moments when it is trying to match the results it has determined you want. It has no judgment when you're off the beaten track a bit. Two things I've found very useful, 1. Prompting it to number equations, that makes follow on questions much easier. 2, using it as a tool to create latex source for derivations, (GPT 4o & Grok) both do a fine job with very good accuracy, at least for my stuff...
Thinking out loud… It occurs to me there seems to be four "kingdoms" we apply Ai to: [1] Search engines for known facts; [2] Mathematical questions; [3] Art/Literature production; [4] Social questions.
The first two seem worthwhile, largely because of what you mentioned here about begin able to check the work. The third fascinates people but doesn't interest me at all When it comes to art or literature, I don't value something one can produce with a prompt rather than their own creative sweat and blood.
The fourth, at least in the foreseeable future, seems very, very dangerous to me.
I totally agree! The fourth is very arbitrary. Also note that the social sciences academic literature has a substantial proportion of non-reproducible studies. It is therefore impossible to logically assess and analyze. This is why AI generated social science papers have made it through peer review.
Your experiences match mine perfectly. Probably the most disconcerting ones for me is its enthusiastic encouragement when I've gone off track and the "miracle occurs here" moments when it is trying to match the results it has determined you want. It has no judgment when you're off the beaten track a bit. Two things I've found very useful, 1. Prompting it to number equations, that makes follow on questions much easier. 2, using it as a tool to create latex source for derivations, (GPT 4o & Grok) both do a fine job with very good accuracy, at least for my stuff...
Thinking out loud… It occurs to me there seems to be four "kingdoms" we apply Ai to: [1] Search engines for known facts; [2] Mathematical questions; [3] Art/Literature production; [4] Social questions.
The first two seem worthwhile, largely because of what you mentioned here about begin able to check the work. The third fascinates people but doesn't interest me at all When it comes to art or literature, I don't value something one can produce with a prompt rather than their own creative sweat and blood.
The fourth, at least in the foreseeable future, seems very, very dangerous to me.
I totally agree! The fourth is very arbitrary. Also note that the social sciences academic literature has a substantial proportion of non-reproducible studies. It is therefore impossible to logically assess and analyze. This is why AI generated social science papers have made it through peer review.
Good point.
There's a bug in Riemann Zeta theory. Pleiades/Springer Nature:
Lobachevskii J of Mathematics, 46(3), 1266-1270, 2025.
I am not interested in “likes ” nor interested in dopamine.
"Arriving at the answer it expects to find."
"Giving you the answer it thinks you want to hear."
That describes all political leaders and the vast majority of undergraduate college students.