Scivillage.com Casual Discussion Science Forum
When You Trust AI - Printable Version

+- Scivillage.com Casual Discussion Science Forum (https://www.scivillage.com)
+-- Forum: Science (https://www.scivillage.com/forum-61.html)
+--- Forum: Computer Sci., Programming & Intelligence (https://www.scivillage.com/forum-79.html)
+--- Thread: When You Trust AI (/thread-19245.html)



When You Trust AI - Zinjanthropos - Nov 23, 2025

Posed to AI…..Is it wise to ask AI to help with a rules exam?

Have couple friends who, to keep their eligibility card, had to write an online exam on the sport they referee. They decided to attempt exam together one night. The passing grade was 90%, 50 situational rules questions all yes or no. It was an open book exam but instead of consulting the rule book they decided it would be easier to just ask AI each test question as they appeared on their monitor screen. As a result they both failed with only 76%. Fortunately all not lost as they get one chance to rewrite next week.

Prompted me to ask the question in the opening line of this post. AI said it is unwise because of the ethics (cheat) involved and people wouldn’t learn or understand by asking, and added this….

Quote:. Inaccuracy and over-simplification: AI can produce incorrect or overly simplistic answers, and it's essential to critically evaluate and verify any information it provides.

Their opinion of AI has changed…lol


RE: When You Trust AI - Syne - Nov 23, 2025

Yeah, you should only rely on AI on topics you are already pretty familiar with. It can be good at compiling/distilling information and opinions (pros & cons), but it is very prone to factual errors. I've had it try to mislead me on a large variety of subjects (programming, mechanics, etc.), often just because it assumes a context not included in the question or doesn't limit its answer to the specific question asked.

Being language models, they prioritize sounding human, complete with errors, over being accurate.

If you know the subject fairly well and really push back on information you don't think sounds valid, you can eventually get useful answers. But it takes so much patience that it often belies the productivity and convenience of using AI.