My Problem with AI!

[Qapla]

 

1 hour ago, Mike047 said:

Some use it rather than theocratic publications, this is a poor option for personal study.

Edited Monday at 06:39 PM by Qapla

[hybridxer0]

On 7/6/2026 at 9:55 AM, Mike047 said:

Like other problems, it often comes down to whether it's a tool you can control. Some use it rather than theocratic publications, this is a poor option for personal study.

The main issue, and one I'm uniquely qualified to speak on... is news/media does a very poor job on educating the gen pop on what AI actually is. Large Language Models are meant to be used as tools in their current capacity to use in tandem with other systems, and definitely should not be considered as primary sources. As some state, it is more akin to a search engine (though a bit too simplistic a comparison). There is logic happening on the back end, and they can "learn" by providing more correct information to replace incorrect/bad information. 

The problem that most don't consider is that none of the major LLMs are open-source, or able to be edited responsibly -- there is always a company with an agenda. I use several LLMs every day for various tasks and some it handles exceptionally. There are so many use cases that go beyond having an encyclopedia in your pocket, that I feel people using them in the wrong way miss out on. There are ways technology can benefit day-to-day life, but the responsibility lies in the user to utilize the tools properly. 

One thing I will say to note, using an LLM to search the library for either concepts or specific phrasing has been a life-saver for personal study as the official search for the JW Library is severely lacking and needs an overhaul.

[Mike047]

5 hours ago, hybridxer0 said:

One thing I will say to note, using an LLM to search the library for either concepts or specific phrasing has been a life-saver for personal study as the official search for the JW Library is severely lacking and needs an overhaul.

This is a useful point you make, Richard, and as long as AI is used in tandem with our own source material, it can certainly be helpful. 

[Qapla]

Here is an example of how AI can be useful:

 

We have been working on my daughter's van. We had to replace the electric cooling fan. After getting the van put back together, we found that the cooling issues went beyond the burned-out fan. Her van has a "captive cooling system" which means, it does not have a radiator cap - instead, the reservoir bottle is pressurized while the vehicle is ion operation and the cap seals tightly to hold the pressure.

 

The reservoir bottle is plastic, and we discovered it has cracks in the bottom, so we had to replace that, too. When we got everything back together - again, we found there is also a small leak near the top of the radiator. OK - we can just use some 'stop leak' ... that's what we used to do it when I was younger.

 

So, I went to the AI search and asked what the best stop leak is to use for this van. Of course, to do so, I specified what the van is. Well, the AI directly told me NOT to use stop leak in that vehicle due to the captive cooling system and a few other reasons. Seems I could do serious and possibly irreparable damage by using the stop leak.

 

That seemed a bit odd and/or extreme - so, I used another AI and asked that one about the stop leak. I got the same type of response. It also told me not to do it unless it was strictly an emergency and we were just trying to limp it off the road or get it back home.

 

So, I asked a mechanic I know. An actual human mechanic who makes his living fixing cars. He verified what the AI's told me.

 

We have a new radiator ordered and should be able to install it next week.

 

In this case, the AI did not give us wrong information, hallucinate or make stuff up. It actually saved us money and headaches.

[Mike047]

And that, John, is where it is very helpful, having had a practical issue to sort out not long ago.