Don’t ask ChatGPT about PLCC44 sockets

[tggzzz]

If you ask ChatGPT a limited scope question with a verifiable way to test its answer, it's a very powerful tool.

Reasonable, but nowhere near as revolutionary as people claim.

More importantly, most people won't understand that principle, and in most cases there won't be a way to verify a response (!answer ), and even of there was a way, most people wouldn't bother. Even "intelligent educated" people like.lawyers.

On that basis, LLMs are a "dangerous" tool.

[SiliconWizard]

Yes. There's also a more fundamental issue, behind its use, even if one could assume it always "answered" correctly (which is still far away).

This will tend to make learning "obsolete". People learn to be able to figure out complex things. They spend years learning, sometimes. And they develop an understanding that no quick answer could give them.
These LLM tools promote the "quick answer" approach to every question, rather than learning. People just want a quick answer to some question, and move on.
The consequences are major.

To make a parallel, imagine that instead of going to school and then to uni, for years, you just asked a professor some question when you had one, took the answer, and moved on. Because this is the model we are heading towards.

And let's take EE as an example: many of us have gotten into EE because at some point in the childhood they wanted to understand how some device worked, then another, and then they figured they had to learn.

Learning makes you an independent thinker, over time, and able to devise new things. Just playing questions/answers makes you a fully dependent pawn. What do you want to be?

[tggzzz]

All valid, but the points are scarcely unique to LLMs.

I count myself lucky that going through university I had to hand write what the lecturer was saying.

Yes, I know the old joke that a lecture is where notes are copied from the lecturer's paper to the students' paper without going through the minds of either. My handwriting was so bad that I recopied the notes within a couple of hours - and that enabled me to make them coherent and legible, and forced me to work out what I hadn't understood. Made revision easier!

Being given typed handouts would have avoided all those benefits.

[SiliconWizard]

It's a well-known fact that note-taking helps with learning.

[vk6zgo]

I just asked ChatGPT about the mapping between the package pins and the pins of a THT socket. In the first reply, the counting started at the corners. After I corrected it several times, the text became correct, but the ascii drawing still had the numbering starting at the corners. No way to fix it. The results for mapping to the THT were very confusing. An "intelligent" answer could have been a recommendation for an SMD socket or using a different package.

If you go back in a couple of days, ChatGPT will have forgotten your corrections & will make up some other nonsense answer, or as in my case say they don't have an answer.

My original test was to ask it about a "Carpenter relay", a relatively well known, but not all that common device in years past.
The answer was a lot of waffle about thermal relays.

When I corrected it, its next effort was basically what I had said, rephrased & as I pointed out above, my corrections were lost when checked some time later, & Chat GPT "Had no information on Carpenter relays".

[vk6zgo]

There's a tendency when one gets older to start ignoring everything said by younger people because they don't have the same years of wisdom that you do and they get things totally wrong.   
This is a fatal mistake in life, and it's the same with asking AI things.

You should ask AI a lot, but it should be obvious that, if you ask about something obscure that it didn't get much training on, the answer is going to be pretty terrible.
The same as it is when talking to a human with limited knowledge of a subject.
AI is not a tool, tools do one job well, AI is an entirely new thing.
The main issue is that AI can't tell when it doesn't know enough to answer a question accurately, so it's confident about everything including stuff it's wrong about

Interestingly, the first time I asked it a question , it told me a lot of drivel, but the next time I asked the same question a few days later, it said it "didn't have any information".
The thought just occurred to me, maybe it recognised my nickname & decided I was a troll!

[hneve]

I had a similar issue with ChatGPT trying to figure out PLCC44 pinouts. It got pretty confusing because it kept messing up the numbering on the ASCII diagrams. I ended up finding better info in datasheets and electronics forums. Sometimes those sources are just more reliable for these details.

[tggzzz]

I had a similar issue with ChatGPT trying to figure out PLCC44 pinouts. It got pretty confusing because it kept messing up the numbering on the ASCII diagrams. I ended up finding better info in datasheets and electronics forums. Sometimes those sources are just more reliable for these details.

Sometimes?!

Under what circumstances would a statistical parrot be more reliable than a definitive source?

[SiliconWizard]

I had a similar issue with ChatGPT trying to figure out PLCC44 pinouts. It got pretty confusing because it kept messing up the numbering on the ASCII diagrams. I ended up finding better info in datasheets and electronics forums. Sometimes those sources are just more reliable for these details.

Sometimes?!

Under what circumstances would a statistical parrot be more reliable than a definitive source?

In a world where people can't tell the difference.

[SiliconWizard]

Interestingly, the first time I asked it a question , it told me a lot of drivel, but the next time I asked the same question a few days later, it said it "didn't have any information".

This is relatively common with chat bots. I've experienced that with commercial support chat bots as well. One day I would ask a question and would get a related answer (albeit very generic), the next day I would re-ask the same thing and would get something like: "Ok. What exactly would you like to know?". And I would rephrase, and it would get stuck in an endless loop of replying this again and again.
Either it's a bug, or it's by design. Maybe the chat bots just reply that kind of stuff by default when the LLM server is too busy, or something.