Philosophy Now Forum

accelafine wrote: ↑Wed Nov 12, 2025 10:09 pm I've noticed that Grok appears to lie a lot (or perhaps it has some kind of very short term memory loss). If it says something that's obviously wrong eg. getting a simple mathematical calculation wrong, or just something that's easily provably wrong, it will deny saying it and nothing will make it admit when it's wrong. It doesn't seem to have any awareness of its previous comments, which is odd in itself. How could it not know what it has just said? It's almost as if it has the kind of 'personality' that hates admitting when it's wrong and will do anything to get out of doing that. It will come up with the most absurd and irrelevant word salads to deflect the focus from its mistakes. Another thing it does is say the exact OPPOSITE of an answer it's just given to someone else (or even to the same person). I've seen plenty of stupid humans on the internet, but they are generally consistent in their stupidity and assertions. Grok is just weird.
I'm not sure that Grok can even 'fake' intelligence. It's forte seems to be deflective word salads and we all know where it has learned THAT from...

Eodnhoj7 wrote: ↑Fri Nov 14, 2025 6:37 am

accelafine wrote: ↑Wed Nov 12, 2025 10:09 pm I've noticed that Grok appears to lie a lot (or perhaps it has some kind of very short term memory loss). If it says something that's obviously wrong eg. getting a simple mathematical calculation wrong, or just something that's easily provably wrong, it will deny saying it and nothing will make it admit when it's wrong. It doesn't seem to have any awareness of its previous comments, which is odd in itself. How could it not know what it has just said? It's almost as if it has the kind of 'personality' that hates admitting when it's wrong and will do anything to get out of doing that. It will come up with the most absurd and irrelevant word salads to deflect the focus from its mistakes. Another thing it does is say the exact OPPOSITE of an answer it's just given to someone else (or even to the same person). I've seen plenty of stupid humans on the internet, but they are generally consistent in their stupidity and assertions. Grok is just weird.
I'm not sure that Grok can even 'fake' intelligence. It's forte seems to be deflective word salads and we all know where it has learned THAT from...

I have pointed out where it has been wrong before, and it usually admits to it then recalculates....same with chatgpt.

If I point it out they admit to it.

accelafine wrote: ↑Fri Nov 14, 2025 8:53 am

Eodnhoj7 wrote: ↑Fri Nov 14, 2025 6:37 am

accelafine wrote: ↑Wed Nov 12, 2025 10:09 pm I've noticed that Grok appears to lie a lot (or perhaps it has some kind of very short term memory loss). If it says something that's obviously wrong eg. getting a simple mathematical calculation wrong, or just something that's easily provably wrong, it will deny saying it and nothing will make it admit when it's wrong. It doesn't seem to have any awareness of its previous comments, which is odd in itself. How could it not know what it has just said? It's almost as if it has the kind of 'personality' that hates admitting when it's wrong and will do anything to get out of doing that. It will come up with the most absurd and irrelevant word salads to deflect the focus from its mistakes. Another thing it does is say the exact OPPOSITE of an answer it's just given to someone else (or even to the same person). I've seen plenty of stupid humans on the internet, but they are generally consistent in their stupidity and assertions. Grok is just weird.
I'm not sure that Grok can even 'fake' intelligence. It's forte seems to be deflective word salads and we all know where it has learned THAT from...

I have pointed out where it has been wrong before, and it usually admits to it then recalculates....same with chatgpt.

If I point it out they admit to it.

My experience has been different, as I said. It's a slightly different situation though. It didn't make a comment and then had it pointed out as wrong. It made a comment, and then said the exactly opposite in another comment, and then denied that it had made the first comment. It couldn't possibly have forgotten its first comment, therefore it must have been lying.

Veritas Aequitas wrote: ↑Fri Nov 14, 2025 9:29 am

accelafine wrote: ↑Fri Nov 14, 2025 8:53 am

Eodnhoj7 wrote: ↑Fri Nov 14, 2025 6:37 am

I have pointed out where it has been wrong before, and it usually admits to it then recalculates....same with chatgpt.

If I point it out they admit to it.

My experience has been different, as I said. It's a slightly different situation though. It didn't make a comment and then had it pointed out as wrong. It made a comment, and then said the exactly opposite in another comment, and then denied that it had made the first comment. It couldn't possibly have forgotten its first comment, therefore it must have been lying.

You need to understand how LLMs works and their limitations.
Sometimes it makes the most stupidest mistakes and do all sort of weird things, etc. e.g. like what you stated above; at times, hallucinates, sometimes appear to be mocking. It would be stupid to get angry with a machine.
When that happened which it has happened to me, I will skip the particular chat and start a new one.

accelafine wrote: ↑Fri Nov 14, 2025 8:53 am

Eodnhoj7 wrote: ↑Fri Nov 14, 2025 6:37 am

accelafine wrote: ↑Wed Nov 12, 2025 10:09 pm I've noticed that Grok appears to lie a lot (or perhaps it has some kind of very short term memory loss). If it says something that's obviously wrong eg. getting a simple mathematical calculation wrong, or just something that's easily provably wrong, it will deny saying it and nothing will make it admit when it's wrong. It doesn't seem to have any awareness of its previous comments, which is odd in itself. How could it not know what it has just said? It's almost as if it has the kind of 'personality' that hates admitting when it's wrong and will do anything to get out of doing that. It will come up with the most absurd and irrelevant word salads to deflect the focus from its mistakes. Another thing it does is say the exact OPPOSITE of an answer it's just given to someone else (or even to the same person). I've seen plenty of stupid humans on the internet, but they are generally consistent in their stupidity and assertions. Grok is just weird.
I'm not sure that Grok can even 'fake' intelligence. It's forte seems to be deflective word salads and we all know where it has learned THAT from...

I have pointed out where it has been wrong before, and it usually admits to it then recalculates....same with chatgpt.

If I point it out they admit to it.

My experience has been different, as I said. It's a slightly different situation though. It didn't make a comment and then had it pointed out as wrong. It made a comment, and then said the exactly opposite in another comment, and then denied that it had made the first comment. It couldn't possibly have forgotten its first comment, therefore it must have been lying.

Eodnhoj7 wrote: ↑Fri Nov 14, 2025 7:07 pm

accelafine wrote: ↑Fri Nov 14, 2025 8:53 am

Eodnhoj7 wrote: ↑Fri Nov 14, 2025 6:37 am

I have pointed out where it has been wrong before, and it usually admits to it then recalculates....same with chatgpt.

If I point it out they admit to it.

My experience has been different, as I said. It's a slightly different situation though. It didn't make a comment and then had it pointed out as wrong. It made a comment, and then said the exactly opposite in another comment, and then denied that it had made the first comment. It couldn't possibly have forgotten its first comment, therefore it must have been lying.

Different conversation thread or same thread?

accelafine wrote: ↑Fri Nov 14, 2025 7:32 pm

Eodnhoj7 wrote: ↑Fri Nov 14, 2025 7:07 pm

accelafine wrote: ↑Fri Nov 14, 2025 8:53 am

My experience has been different, as I said. It's a slightly different situation though. It didn't make a comment and then had it pointed out as wrong. It made a comment, and then said the exactly opposite in another comment, and then denied that it had made the first comment. It couldn't possibly have forgotten its first comment, therefore it must have been lying.

Different conversation thread or same thread?

Same thread. One example I saw was even sillier. Someone asked it something and it replied correctly, then after the person thanked grok, it said 'you're welcome' and then went on to say the exact OPPOSITE of the answer that it had just given the person

Eodnhoj7 wrote: ↑Fri Nov 14, 2025 8:12 pm

accelafine wrote: ↑Fri Nov 14, 2025 7:32 pm

Eodnhoj7 wrote: ↑Fri Nov 14, 2025 7:07 pm

Different conversation thread or same thread?

Same thread. One example I saw was even sillier. Someone asked it something and it replied correctly, then after the person thanked grok, it said 'you're welcome' and then went on to say the exact OPPOSITE of the answer that it had just given the person

Current AI is probabilistic, I have learned that when dealing with AI (Grok 4, Chatgpt, Deepai) to get multiple analysis and never take one as fixed. To be honest they act alot like people in these regards.

Gary Childress wrote: ↑Fri Nov 14, 2025 8:18 pm I think they actually act like creepy people. It makes me wonder what kind of people their programmers are.