Please don't trust AI blindly

It often gives misinformation like so:

i just read that ai exhibits concerning behavior by making up data and denying mistakes

'But I am too lazy to, it is too hard to, and it takes too much effort to think for myself.' 'If I get an answer from someone else, then I can blame someone else if it is incorrect.' 'If others blindly follow AI, then so should I; mob mentality all the way.' --- The average human.

Don't trust AI blindly. Don't trust youtube blindly. Don't trust your favorite news source blindly, etc...

At the same time, "Who was helping Daniel Nairoditsky" is an awfully vague question. Who was helping him with what? If you want to get the most out of AI, you have to provide context and be specific. BTW, isn't it true (and ridiculous) that Kramnik claimed to to have helped him before his death?

But yes, I wish AI was a bit less confident in its answers. It should have responded with "what do you mean by helping?" and then "there's not a lot of information about this online".

"Who was helping Daniel Nairoditsky with his depression" gets a better chatgpt response. See below. I still wouldn't take it at face value but I'd use it as a starting point for my research if I were interested in this.

Publicly, the clearest evidence is that people close to Naroditsky in the chess community were supporting him, especially Daniel Rensch of Chess.com. Rensch later wrote that Naroditsky had told him his reduced commentary work was about his mental health, and that they had many phone calls, texts, and emails about what he was going through. Chess.com
There are also reports that his Charlotte chess community and friends were actively trying to help him, including in his final hours. US Chess
What I have not found is reliable public evidence naming a specific therapist, psychiatrist, or other mental-health professional treating him for depression.

I have caught AI repeatedly making mistakes even when solving problems. And that isn't conjecture or hubris -- I've gotten the AI to admit it erred and REVERSE its original answer. I had quickly sensed that it was in error because of my own, human mathematical intuition.

And this has happened more than once.

Lawyers, and even judges, have been tricked when AI just made up a case in order to make an argument!

Indeed, that tendency makes me fear that AI is not at all conservative. I suspect it would vote for silly candidates, too.

AI has to be used, at least in some contexts, carefully and with real understanding of its capacity for error, and how to minimize the possibility of error. I fear many people will get wrong answers who will have no idea to challenge them, but just accept them uncritically and become cemented into bad positions.

On the other hand, in some fields AI is wonderful. But a lot of that work is not done by the AI that most of us readily use.

Can't say they didn't warn us:

AI talks a lot of schitt about chess, that's for sure

check gemini.com not the slop overview

@fallboss007 said ^

check gemini.com not the slop overview

gemini.com is a stock market site

@qpalzm123456 said ^

It often gives misinformation like so:

image.png (https://image.lichess1.org/display?fmt=png&h=0&op=resize&path=Otxw1m52CdMW.png&w=864&sig=cea267aa22df4354cd0f9f8d42f31cd048b56de1)

dosent work for me though.