Dropping threads due to posts about GPT hallucinations?

[Nominal Animal]

Am I the only one here who has started to completely ignore threads when members post what their favourite GPT hallucinates about the topic?



To me, it's like listening to parents excitedly describe the opinions their 4-year old "precocious genius" has about sustainable business practices, or technical details of long-term tax planning.  Interesting to parents discussing their children, but decidedly anti-interesting when discussing actual technical stuff.

I suppose it is not any different to dropping threads because of tone or political sidetracks, though.

Edit: Fixed typo in the darn title.  it's "due to", not "dues to".  The latter are fees.

[mendip_discovery]

I feel you.

I am not sure of the motivations for just asking the question asked to the LLM bots. If I wanted that I would have asked one myself. I wonder about the motivations as some feel it should be part of everyday life and want to continue the conversation on how useful it is. On the other side you have some that like to do that "would you like me to google that for you" type moments.

I have tried to use it to write code but I don't like it as it made me feel awkward about plagiarism. As I don't like data lacking a source.

[pcprogrammer]

I ignore basically everything that hints about any of the "AI" stuff. I rather think for myself. 

Also filter out to long to read posts due to concentration issues. I like short and concise sentences instead of large paragraphs.

[Nominal Animal]

I like short and concise sentences instead of large paragraphs.

You monster! 

Yeah, I much prefer concise and to the point myself.  My verbosity is an affliction, a burden, I'd like to learn out of.


Those "I asked GPT, and it said" posts are, to me, the same as if someone did a search on StackOverflow/StackExchange/Reddit, and posted what they wrote there, without any kind of an analysis whether it makes any sense or not.

One reason my programming and math related posts are so long is that I don't like to tell others to simply believe or trust what I say; I want to show how what the solution is based on, how to verify it for yourself, and how to use the same search pattern to find solutions to similar problems.  I'd hate to be considered "an authority" on any subject, someone to be trusted and not questioned; I much prefer the role of a "guide", who often says "trust if you must, but always verify".

In that light, GPT answers really are just hallucinations, with an accompanying "trust me bro, I read the whole internet once" assurance.

[shapirus]

My verbosity is an affliction, a burden, I'd like to learn out of.

I find it helpful, for a starting point, to begin making paragraphs smaller. I know it's hard, but it seems to work :)
The rest will catch up eventually.

[pcprogrammer]

I like short and concise sentences instead of large paragraphs.

You monster! 

I knew that it was going to bite me. 

[rstofer]

The train is leaving the station, it's best to get on board.  VisiCalc, considered the first spreadsheet, made Apple.  By '79, Apple II hardware couldn't  compete with PCs but VisiCalc came along and guaranteed it's survival.

Now comes AI and, eventually, it will be a massive job killer except for those who know how to apply it.  It's pretty clear that fast food won't survive the new $20/hour mandate in California unless the hours are reduced.  Look for FULL automation in the VERY near future.  AI is very likely involved in this transition.  Speech recognition is an obvious starting point.  There's not much of that in my neighborhood but it's coming soon!

[Nominal Animal]

The train is leaving the station, it's best to get on board.

When the train is going to the La-La Land where facts don't matter, math and logic are old-fashioned, and surface appearance and ease of use is all that matters, I prefer to go elsewhere, thank you.

None of the current "AI" can combine logic with their generated output, and nobody has even a functional theoretical model on how that could be done: it's like trying to bolt a hinge to a cloud.  Until that can be done, or we limit to models trained with human-vetted data, that train is definitely running on LSD.

[soldar]

Also filter out to long to read posts due to concentration issues. I like short and concise sentences instead of large paragraphs.

I, OTOH, take my time to do my best to explain what I am trying to ask or say and I hate it when half the posts misunderstand the question and are answering to something unrelated just because people are to lazy to focus and read carefully.

And I hate it when someone posts a question so vague and so badly explained that people have to come in and ask a hundred questions from the OP just to find out what he is asking. If you are so lazy and put so little effort into your question I am not going to bother myself. 

I used to post snarky comments but now I try to just ignore them and move on.

[RJSV]

   Had to go use 'google' to describe the means ng, in more exact terms:
   GPT. = Generational Pre-Trained

   I think you are saying that it's often a 'junk' output, in ignorance of even simple human subtleties.   Some places use phrase 'the unsaid' meaning that there is lots of hidden nuance.   (Am I right, so far ?).
   Like, for example, a friend who, unfortunately STINKS, and folks kind of dance around the issue, rather than making literal comments.  That's going to include some non-verbal things like raising eyebrows, or winking, when Uncle Frederick is mentioned.  That kind of nuance gets completely lost, I would think, by any mechanical 'GPT'...

   Myself, I'm not far enough along to start excluding threads, but I see what you mean, about assessing the worth, of spending time.

As to posting brevity, that's good.   But some exceptions, when diving deep into subjects.
Loved the posts by Nominal Animal on the recent subject of 'Randomness' and math - coding details....THAT would be the exception.
Reading that detailed run-down, on random outputs and math, I realized just that could occupy a week, in some university oftware course

[magic]

I asked ChatGPT what it thinks about your thread, do you want to know?

[Nominal Animal]

I think you are saying that it's often a 'junk' output, in ignorance of even simple human subtleties.

Not really.. You see, "hallucination" is the correct technical term that the generational "AI" developers use to describe this.

Generational 'AI's are the technological equivalent to a lyrebird.  They combine and mix their training data to match the prompt(s), producing very good fidelity and surface appearance, but zero understanding of the content or logic.  Unlike a lyrebird, they don't just mix and match at random, but ensure continuity and matching of the whole output.  That is all they are.  Amazing, but not "intelligent" by any stretch.  Thus, it will happily claim that it has proof that 1+1=3, or that it is a good idea to use glue to ensure your cheese sticks to your pizza dough.

"Hallucination" describes the situation where the transformer produces output by "incorrectly" mixing its training data, producing believable and at the surface acceptable output, but completely without a basis in its training data (even if it claims so; those claims themselves are the same hallucination).  Thing is, it is really not "incorrectly" mixed; it always does that, because it has no understanding or logic.

I asked ChatGPT what it thinks about your thread, do you want to know?

I'm more interested in what it suggests one does when one wants to smooth the corners in ones stone sphere.  Or how to remove one corner from a cube and still get a regular platonic solid.

[nctnico]

The train is leaving the station, it's best to get on board.

When the train is going to the La-La Land where facts don't matter, math and logic are old-fashioned, and surface appearance and ease of use is all that matters, I prefer to go elsewhere, thank you.

None of the current "AI" can combine logic with their generated output, and nobody has even a functional theoretical model on how that could be done: it's like trying to bolt a hinge to a cloud.  Until that can be done, or we limit to models trained with human-vetted data, that train is definitely running on LSD.

The biggest danger I see is that people are using AI to create things they don't understand. Last week I was in a meeting with a very junior software engineer who had used ChatGPT to write a bit of code. This code does what is is supposed to do, but what if it doesn't? Or what if the code needs changes? Then we're screwed and I likely find myself doing a job which was not assigned to me.

IMHO what is good about ChatGPT (or better put, the technology behind it) is that it can take in a lot of data and extract useful excerpts from it. There is so much information hidden in archives and libraries that it is impossible for a single person to even get an overview of the information in a lifetime even for a single archive / library. IOW: the application for AI I see is to organise information and point people towards the relevant information sources they need.

[Siwastaja]

IMHO what is good about ChatGPT (or better put, the technology behind it) is that it can take in a lot of data and extract useful excerpts from it.

ChatGPT is just a "better Google". People are so excited about it because classic search engines, for some reason I can't understand, not only stayed in stone age but actually regressed. For example, the Google Search (the website / text search, not talking about images, maps etc.) was in its peak technical performance somewhere in 2003. Since then, it was purposely made worse and worse, with the aim to return maximum number of search results, instead of high quality results. The starting point in late 1990's was OK-ish: search engines returned web pages which contained the words in the prompt. Mediocre, but very simple, and you usually found something about the actual subject you wanted to know within 50-100 first results.

In early years of 2000's, Google made some algorithmic improvements invisible to end users which improved results and brought relevant information to the first 10 hits, but then downhill started soon, and still in 2020, if you need any specific information, it's simply impossible to find using Google.

For two full decades as of now, Google's primary use is to allow people to type "face bok", get 1000000000000000000 results and find facebook.com as the first result. For any more specific research, you can't find anything.

Then came ChatGPT so that you can ask a specific question, and get random and mixed quality answers, it's like Google in 1999 just formatted in more verbose bullshit boilerplate structure, plus of course the fact that you have to separately ask for the references, and if lucky, get something. It's also pretty good at finding what you actually wanted to know, even if your keywords are not exact. Similar to what Google researched on their algorithms in very early years of 2000's.

But it allows finding information from the Internets again, like it was 1999 all over again. And the excitement is the same. Just like people were overexcited in 1999 and Google became a business giant thanks to its good start with search engines, we are now seeing the same hype about "AI". The end result will be the same all over again: modest improvement, stall of development, regression, and disappointment.

For many years already, AI has been said to "improve dramatically" "within just a few years" and "take all the jobs". Yet, nothing has happened: ChatGPT nor its competitors have not improved at all (some feel they have already regressed). People like rstofer play this "oh but within next two years it will happen!" game for a few more years, and then not only forget about the whole thing, also forget about what they said.

And then comes the next hype, and the cycle repeats.

[RJSV]

Magic:
   By all means;  Please.   I wouldn't mind even if
the Chat review of my post wasnt flattering !
   Curious, and not offended.
   Can you copy the review, of my previous post (Yes, I'm that vain,  lol).

[DimitriP]

Am I the only one here who has started to completely ignore threads when members post what their favourite GPT hallucinates about the topic?

Nope. There are at least two !!

Those "I asked GPT, and it said" posts are, to me, the same as if someone did a search on StackOverflow/StackExchange/Reddit, and posted what they wrote there, without any kind of an analysis whether it makes any sense or not.

I miss using mIRC  where you could "slap <nickname> with a large trout". 

[shapirus]

I miss using mIRC  where you could "slap <nickname> with a large trout".

Around. You forgot the "around".

[soldar]

ChatGPT is just a "better Google". People are so excited about it because classic search engines, for some reason I can't understand, not only stayed in stone age but actually regressed.

Speaking of search engines ... you can go to eBay or other commercial places and type something and get results for what you want .... except Amazon. For reasons I do not know they give you mostly results totally unrelated to your search. I have no idea why but I find it frustrating. In the end I end up looking for products elsewhere and sometimes return to order at Amazon if they have good price and conditions. I don't get it. This is not artificial intelligence, it is artificial stupidity.

[Bud]

. For reasons I do not know they give you mostly results totally unrelated to your search. I have no idea why but I find it frustrating.

Same reason when department/groccery stores periodically re-arrange products to different aisles. They want you to spend more time in the store and look at more products, chances are you will grab more stuff while looking for what you came for.

[Nominal Animal]

IMHO what is good about ChatGPT (or better put, the technology behind it) is that it can take in a lot of data and extract useful excerpts from it.

Yep, I agree with that.  The technology itself is very interesting (albeit limited in ways ordinary people seem to have a hard time understanding –– like the "not intelligent" part); right now, it is just used to produce more crap.

(I like Sci-Fi, and the amount of GPT-generated pap posing as sci-fi stories out there annoys me to all hell.  It doesn't help that a sci-fi story will have a hard time getting published unless the protagonist is a non-male, preferably non-binary POC, fighting against light-skinned evil males.  The storyline doesn't matter; the group membership and "depth" of characters does.  There are legal free sci-fi audiobooks on Youtube, like librivox, but they're easier to find elsewhere (on LibriVox.org, for example).  Try to find one on Youtube, and you're swamped with "GPT-aided" HFY and creepypasta crap; or "AI-voice" -read ebooks that fool you for a few seconds until you notice it doesn't do section/chapter breaks, and pronounces most non-English names hilariously wrong.)

For example, the Google Search (the website / text search, not talking about images, maps etc.) was in its peak technical performance somewhere in 2003.

I aggressively agree.  I miss the days when I could use +term1 -term2 to look for pages that contain term1 but not term2.  Now, it suffices that the page is similar to or references another page that at some point did contain those terms, or was itself similar to or referenced another page that at some point did contain those terms, and so on ad infinitum.

The Youtube "search" is an even worse offender.  It does not even try to match the results to the query.  Recently, I did a search on "Neal Asher" (the author of Polity sci-fi novels, quoted to hopefully get exact matches).  Because he has a video titled "Raki-Making on Crete" from 13 years ago, the search results include all sorts of videos related to Crete, Raki, and making (stuff).  And the first match is a year-old no-views "Asher Neal - appearance" video.

The way I can find anything on Youtube anymore is to find the author/uploader, and check their Videos list.  The search is absolutely useless for finding a specific video.  And all the good ones I like and want to see tend to be "down-rated" or shadowbanned by the search algorithm.  I am just not interested in the premasticated massproduced pap Google/Youtube/Alphabet/Advertisers want to push.

And then comes the next hype, and the cycle repeats.

I remember the time when VR and virtual reality environments were supposed to change the way we access Internet.  For a while, it was 3D TVs.

I think the train goes in a circle, with the conductor offering free LSD to everyone aboard, as long as they look ecstatic about being on the Very Train, Not Left Out.

[RoGeorge]

There was a movie with a train running in circles

[pcprogrammer]

. For reasons I do not know they give you mostly results totally unrelated to your search. I have no idea why but I find it frustrating.

Same reason when department/groccery stores periodically re-arrange products to different aisles. They want you to spend more time in the store and look at more products, chances are you will grab more stuff while looking for what you came for.

We have psychologists, psychiatrist and psychoanalysts to thank for that. Seems to work well on most people, otherwise the shops won't bother, but to me it is frustrating as hell. Same as commercials on the TV, where they repeat the same one several times in the same commercial block as if every body suffers from dementia and needs to be reminded of something every minute or so. Also a well liked format in several TV programs, where they address a subject, not to finish it directly, change over to some other story, then come back to the first subject, to repeat about half of it and so on. 30 minutes of TV with only maybe 10 minutes worth of actual data. The rest is repeats and adds.

On google search, it takes certain skills to find what you want, and then more skill to filter the results on usefulness, which seems to diminish within the global population by the minute. Might have something to do with all these stupid commercials and programs on TV. You can feel the IQ points slip away when you look at them. 

And it seems that doing research now means just asking questions on a forum and don't bother thinking about it to much. But indeed what then, if something spouted by some "AI" does not work as intended. How to solve it? Maybe asking another "AI" can help?

We will just have to wait and see where it all leads to.

[SiliconWizard]

This is degenerative AI.

For those who are currently hypnotized by "AI": whatever you do, just try and keep your brain. I promise it will be useful.

[RJSV]

   Somehow,  some way,  when talking about the most premier search engine, (maybe pre-populsr AI),  there naturally has to be some mention of CREEPY Factors.

   Instincts kick in, when there is a stranger sitting out in an unfamiliar car, watching your house (hopefully not, of course).
Or when your email phrases show up in search auto-complete.   Ditto for occasional things spoken.   Don't need a search engine suggesting airplane tickets, when You've just been discussing travel to Missouri, near the cellphone 'listen....' microphone.

   When every last word you've typed, in message to Aunt Martha, gets scrutinized, by your 'mail app'.

   When a 'controversy' erupts over foreign spying via Tiktok evokes a defense, that;
   '...your guys are doing it too'.  on the part of the foreign media defending usual practices.

CREEPY things, and often for free.   (Most services like to charge for what they do).

[tooki]

The train is leaving the station, it's best to get on board. 
…
Now comes AI and, eventually, it will be a massive job killer except for those who know how to apply it.

The train is leaving the station before it’s even been assembled. The wheels are square, the cab doesn’t have a speedometer, the brakes don’t work, and the track gauge is different on the front and back bogies.

That, my friend, is a recipe for disaster for many (most?) applications, and that’s exactly what we are seeing everywhere this half-baked preemie AI has been inserted.

VisiCalc, considered the first spreadsheet, made Apple.  By '79, Apple II hardware couldn't  compete with PCs but VisiCalc came along and guaranteed it's survival.

Released by VisiCorp for the Apple II.