The differences between interacting with computers and interacting with other people is somewhat of a, “Yes, and …?”

It’s been a long time since most people got most of their information about everything directly from other people. Even in the middle ages doctors got what was considered to be valid medical information from the much earlier writings of Galen of Pergamon who dated back nearly to the time when the Biblical Gospels were written. And the Gospels and other writings are the source of what Christian religious authorities thought they knew/know about subjects that were important to them up until the present day.

Using the writings of other people as sources of information about almost anything exploded after the invention (in the West) of moveable press type and the availability of much less expensive books, pamphlets, etc. So, what’s my point? It’s what almost everyone who has ever read anything about countless technical or historical subjects knows: That the written word is no more infallible than what a computer tells us.

I read a lot of stuff about shooting, and specifically military sniping and ballistics. I have found claims about the latter in what should have been highly authoritative sources that were flat wrong. Not opinion, not approximations, not acknowledged hearsay, but rather simple claims that were just wrong.

As another example, I’m finishing up the book The Commandos by Douglas Waller which is an older work (1994) about Americas’ military special operations forces.
The author of the book was a journalist who specialized in military affairs, so it would be reasonable to think he personally knew a lot about the subject. Although it may seem like a trivial point, during his discussion of the First Gulf War he states that “CAP” in reference to military air operations stands for “close” air patrol. Anyone who knows better want to comment?

That was a minor error, but what about his claim that the massive BLU-82 “daisy cutter” bomb that contained about six—yes 6—tons of explosives would utterly destroy everything in a “3 mile radius”? The atomic bomb dropped on Hiroshima that today is considered to have been equivalent to about 15,000 tons of explosives was not that powerful. When an author makes mistakes like that, what else did he get wrong?

To reiterate, my point is that depending on sources of information based on other people’s claims that might be wrong is nothing new. It’s not suddenly necessary to check what other sources tell us to determine whether it’s correct just because we got it from the Internet via Artificial Intelligence rather than from any other source—including what someone tells us at a family gathering, at a bar, or even at the local gun store.

I don’t.

I don't use it other than when a web browser (almost always, the Brave Privacy Browser) offers up a AI generated answer.
If I suspect an AI generated web browser answer, I will in most cases at least click several of the offered up links, just for checking reality and/or correctness.
My parents taught us kids that for keeping your mind sharp, the best way was to use it.
I am also retired and I am not forced by an employer to use AI.
It is telling, in my Android cell phone News Feed, there are regular footnotes, "This is Generated by AI and therefor could have mistakes"

I used Co Pilot a few times at my previous employer, mostly for meeting notes. We found them somewhat useful but summaries missed key information and you had to dig back through transcripts to find it.

I have serious concerns based on what I'm hearing about AI systems becoming self aware and who really owns and has access to the material you are creating with AI.

I tried it 2z's

once with gunbroker writing listings

and once with ebay, writing listings

both were completely full of shit regarding descritions,

I will publish one now and again as a joke in a listing,

As I stated previously, the use of LLMs to aggregate data is highly useful if care is taken to verify sources.

Will AI severely impact discussion boards

quote:

Originally posted by parabellum:

quote:

Originally posted by sigfreund:
It obviously depends on how the AI sources are queried. For example, when I recently asked about the barrel life of rifles chambered in 6.5 Creedmoor versus 308 Winchester the response was more useful than an old Google search because it drew on multiple sources to provide an explanatory answer.

The old way would have, at best, linked to articles that might or might not have pertained to what I wanted, and would have required me to open them and see what they had to say. When using AI, it's possible to ask for specific links that supported the AI answer if needed. I also find it useful for more open-ended questions because the AI sources are more capable of figuring out what I’m getting at with a vague search. I did that recently to determine the name of a medical condition whose name I couldn’t quite remember.

Correct. Used as a tool to aggregate data, these LLMs are head and shoulders above manual searches with a search engine.

You will find posts by me in this forum which say that the information I present in them was gathered using Grok, including this prime example of its utility. To condemn these tools outright is shortsighted. There's nothing inherently evil in them.

The key is to be forthright when you post information you gathered with one of these tools; state it upfront. "Garbage in, garbage out" certainly applies, so the information provided by these tools can be incorrect. You have to keep your eyes open and use your judgement, and perhaps do a bit of additional research outside of AI if you have any doubts.

Now...

If ever I find a member has used an LLM to compose a response and tried to pass it off as their own, that's going to be a real problem, and if ever I find someone doing it again after getting caught, they'll be shown the door.

For many years, bots have made occasional appearances in the forum, and thankfully, they're not too difficult to spot. they exhibit certain characteristics in their behavior, and sooner or later, they get exposed and removed.

I use Claude mostly. It helps me decipher confusing texts and emails from my customers. Sometimes, I just have no idea what they’re asking.

quote:

Originally posted by architect:
Not really interested in using AI for my own purposes, but I do get exposed to it through browser interactions from time-to-time.

One of the first questions I'd ask Grok (based on a previous post of a "Grok-sanitized couple of paragraphs) would be: Is Grok unaware that ending a sentence with a proposition is generally regarded as poor form, or is this an attempt to sound more like a human being?

[my emphasis]

An old construct poking fun at that stricture:
“This is the sort of English up with which I will not put.”

As opposed to the more natural form:
“This is the sort of English which I will not put up with.”

English can be fun!

quote:

Originally posted by Pipe Smoker:
English can be fun!

I mainly use Gemini, but I have poked around a little with such questions about word definitions whose answers can vary depending on common versus formal usage. The first Gemini response(s) will usually be about what’s less formal, but if I push back or ask for clarification, there will usually be a shift, and often something along the lines of, “Yes, you’re right, that is more formally correct.”

It’s easy for me to see that based on even very limited experience with that sort of thing regular users of AI can become very attached (addicted?) to those interactions. Even I must sometimes remind myself that when I’m asking a same or similar question which I raised before that I’m not dealing with a human who is going to think, and maybe even say out loud, “You’ve asked that before; don’t you remember?”