‘Writing’ wrongs in AI

I try to avoid it as much as I can.

AI gets the majority of it’s information by ‘scraping’ the web, mainly scraping Social Media, and we all know how much incorrect cr*p gets posted (published?) on there.

It may get better but at the moment most AI is not discerning enough to discriminate between the good information and the bad information  and so it just serves up the most frequently posted information.

As you have found with some AI answers.
Just because “everybody knows” something doesn’t mean that what they “know” is correct.

Yes, I get that

I’m inclined to try out new options to see how they operate but much like any web search, I never take what’s given as accurate.

I have found some parts useful though to dig into manufacturers info more quickly than standard search.

It's just knowing to double check everything.

Chat without a doubt seems to be the most reliable. The earlier incarnation of Gemini called Bard used to lie through its backside. I won't touch that with a barge pole.

Generally making inquiries of a complicated niche or minority interest through Google just produces blunderbuss type garbage and you could spend a lifetime trying to find the answer if you have the patience to do so. I've found Chat to hit the bullseye first time with a number of arcane questions (one even involving a possible dilemma in general relativity and it told me where I had gone wrong) and I've generally found that inaccurate or off target answers have been due to phrasing the question poorly. With Chat though you can add a correction by giving a more defined context (I use that word) to reframe the question and it generally does the job. HOWEVER, you must verify the answer and Chat gives the sources of the information which you can click on. Interestingly, some of the sources are behind paywalls which makes following up a bit difficult but it shows that Chat can get to places not referred to by some search engines. Sometimes I use it as a foundation for expanding into a broader piece of work. Support or framework for being creative rather than the end product.

I use it sparingly and hardly ever ask it to create images which is very processor/energy intensive.

I have to specify that this is all my personal opinion! 😅

We use ChatGPT at home primarily for research, especially when planning holidays, and I also use on Grammarly to boost my writing confidence. ChatGPT can be really helpful for creating itineraries, but I think it's really important to double-check what it provides and not trust it completely. There have been a few times when it suggested routes that didn’t quite make sense. An example would be that it reccomended a destination which was eight hours off the path! While it wasn't a huge problem since we’re used to taking night trains, it definitely added some unnecessary travel hassle/travel time.

On a different note, I recently asked for advice regarding a health issue, and I was pleasantly surprised by how effective ChatGPT was. It was much better than a visit to my local GP, who didn’t seem to have the time to really help. ChatGPT asked follow-up questions, suggested specific exercises, and encouraged me to report back on my findings. It even provided three possible diagnoses and recommended that I discuss them with a consultant for further testing.

My partner, who works in healthcare, was really impressed with the thoroughness of ChatGPT's responses, especially given my past experiences with the NHS. 

Overall, I think ChatGPT and similar tools can be fantastic support resources, but I always make sure to verify their accuracy myself!