The artificial intelligence MEGATHREAD

[Gadianton]

When you have your own eternal domain, you can set the rules such that AI content is acceptable, if the poster "approves" the content.

Will you then permit posters to constantly complain about your rules, and insist that your rules stack the deck against them - a.k.a. acting the martyr?

You're on the right track. I'd only add that once he allows all of this, the posters complaining will use AI to respond in thousand word responses to everything he says. Of course he won't like it, he wants to be the one using AI on other people not the one getting AI used on him.

[Limnor]

When you have your own eternal domain, you can set the rules such that AI content is acceptable, if the poster "approves" the content.

Will you then permit posters to constantly complain about your rules, and insist that your rules stack the deck against them - a.k.a. acting the martyr?

You're on the right track. I'd only add that once he allows all of this, the posters complaining will use AI to respond in thousand word responses to everything he says. Of course he won't like it, he wants to be the one using AI on other people not the one getting AI used on him.

What seems to matter most is having an immediate reply, not whether it’s coherent or responsive to the actual point. Once others do the same, speed stops serving as a substitute for coherency, and then we see the objection.

[Limnor]

I got it—Godhood means having an answer before understanding the question.

[Physics Guy]

My own idea of when and how it could be okay to use AI on a message board is like my idea of how students are allowed to prepare for their oral exams: if I can’t tell you’re cheating, you’re not cheating.

I don’t just mean that there’s no way to pin the cheating on you. I mean that, whatever you did, to get past me you will have to have done enough actual learning that you legitimately deserve to pass. And in the same way I reckon that simply recognizing when an AI has made a good point concisely, and posting it, is a worthwhile human contribution to a discussion, not really different from tossing in an apt quotation.

I realize that Shades sees it differently. For me, though, the problem is not AI per se, but that what some people seem to want from AI is precisely its bad part—the slick, authoritative-sounding, professionally presented list of bullet points that is totally unreliable in content but has been optimized to sound good.

AI text is annoying because its optimized presentation means that it’s like an over-packaged product. The product itself might be okay, or it might be crap, but it’s wrapped in so much tough plastic that you have to use pliers to get into it, so its only use is to pass it on to someone else, unopened, hoping that they’ll be impressed by the package. Since they probably won’t open it, either, you’ll get credit for giving a gift.

If someone does actually open the package, and find something good, and give that to me, then I don’t mind. They may not have made the thing themselves, but I’m grateful for the work they did with the pliers, cutting through the slick style to expose the substance.

[Marcus]

I realize that Shades sees it differently.

No, he really doesn't. The rule is for our resident troll, not for normal, human thinkers who are interested in discussion.

For me, though, the problem is not AI per se, but that what some people seem to want from AI is precisely its bad part—the slick, authoritative-sounding, professionally presented list of bullet points that is totally unreliable in content but has been heavily optimized to sound good.

That's exactly what MG wants from AI, and exactly why Shades has banned him from using it.

AI text is annoying because its optimized presentation means that it’s like a massively over-packaged product.The product itself might be okay, or it might be crap, but it’s wrapped in so much tough plastic that you have to use pliers to get into it, so its only use is to pass it on to someone else, unopened, hoping that they’ll be impressed by the package. Since they probably won’t open it, either, you’ll get credit for giving a gift.

MG is hoping no one will open it, but only because he uses it to troll and not because he intends it to be actual useful information. He doesn't want credit for a gift, he wants his superiority to reign. The problem (for him) is his use of AI is so blatantly troll-like that no one sees his 'superiority', and no one is fooled.

[Rivendale]

My own idea of when and how it could be okay to use AI on a message board is like my idea of how students are allowed to prepare for their oral exams: if I can’t tell you’re cheating, you’re not cheating.

I don’t just mean that there’s no way to pin the cheating on you. I mean that, whatever you did, to get past me you will have to have done enough actual learning that you legitimately deserve to pass. And in the same way I reckon that simply recognizing when an AI has made a good point concisely, and posting it, is a worthwhile human contribution to a discussion, not really different from tossing in an apt quotation.

I realize that Shades sees it differently. For me, though, the problem is not AI per se, but that what some people seem to want from AI is precisely its bad part—the slick, authoritative-sounding, professionally presented list of bullet points that is totally unreliable in content but has been optimized to sound good.

AI text is annoying because its optimized presentation means that it’s like an over-packaged product. The product itself might be okay, or it might be crap, but it’s wrapped in so much tough plastic that you have to use pliers to get into it, so its only use is to pass it on to someone else, unopened, hoping that they’ll be impressed by the package. Since they probably won’t open it, either, you’ll get credit for giving a gift.

If someone does actually open the package, and find something good, and give that to me, then I don’t mind. They may not have made the thing themselves, but I’m grateful for the work they did with the pliers, cutting through the slick style to expose the substance.

Thanks for reviving an old memory from Wait Wait Don't Tell Me! Tom Bodett.

Bodett described his life as being sold in a clear plastic blister pack, much like cheap goods, with everything labeled and portioned out (e.g., "Blister-pack cigarettes," "Aspiration").

[malkie]

My own idea of when and how it could be okay to use AI on a message board is like my idea of how students are allowed to prepare for their oral exams: if I can’t tell you’re cheating, you’re not cheating.

I don’t just mean that there’s no way to pin the cheating on you. I mean that, whatever you did, to get past me you will have to have done enough actual learning that you legitimately deserve to pass. And in the same way I reckon that simply recognizing when an AI has made a good point concisely, and posting it, is a worthwhile human contribution to a discussion, not really different from tossing in an apt quotation.

I agree with this. Of course, message boards and oral exams work quite differently, but I can see your point. It might be interesting if we could subject posters to oral examination to justify their posts. I'm not sure I would fare well in that environment, based on how inadequate I've felt in the past during oral examinations for language classes.

If you use AI extensively, then you will almost certainly make a good contribution to a discussion - sometimes - perhaps even most of the time. That a good contribution has been made does not necessarily mean that the poster recognised the quality (or conciseness) of the material - they may even have been completely oblivious to the best features of the AI-generated material.

I realize that Shades sees it differently. For me, though, the problem is not AI per se, but that what some people seem to want from AI is precisely its bad part—the slick, authoritative-sounding, professionally presented list of bullet points that is totally unreliable in content but has been optimized to sound good.

AI text is annoying because its optimized presentation means that it’s like an over-packaged product. The product itself might be okay, or it might be crap, but it’s wrapped in so much tough plastic that you have to use pliers to get into it, so its only use is to pass it on to someone else, unopened, hoping that they’ll be impressed by the package. Since they probably won’t open it, either, you’ll get credit for giving a gift.

If someone does actually open the package, and find something good, and give that to me, then I don’t mind. They may not have made the thing themselves, but I’m grateful for the work they did with the pliers, cutting through the slick style to expose the substance.

In an ideal world I would still like to see AI-based content being allowed, but solely as a source - just as if it had been taken from any other resource - book, paper, website - which the poster then subjects to the usual processes of paraphrasing, summarizing, and selectively quoting, then providing the citation that includes at least the name and version of the AI used, date and time, and the exact prompt given. I think this might satisfy your wish for exposure of the substance, along with the skill and understanding of the person wielding the pliers.

Unfortunately, there may be a hidden factor that makes the citation less useful than one might hope. I believe that, as AIs learn during a conversation, they are susceptible to "hints" that may be more or less explicit. This factor may bias their response to a prompt, even if the hints were given earlier in the conversation, or, in some cases in prior conversation sessions.

Possibly, at least with ChatGPT's current public version, you may be able to avoid this sort of bias by asking for all prior discussion on a topic to be disregarded. Without such instruction, ChatGPT may factor in subtextual prompting.

In any case, as far as this board is concerned, for good or ill Shades has made his decision, and, to my mind, has adequately explained his reasoning.

[Whiskey]

I realize that Shades sees it differently.

No, he really doesn't. The rule is for our resident troll, not for normal, human thinkers who are interested in discussion.

For me, though, the problem is not AI per se, but that what some people seem to want from AI is precisely its bad part—the slick, authoritative-sounding, professionally presented list of bullet points that is totally unreliable in content but has been heavily optimized to sound good.

That's exactly what MG wants from AI, and exactly why Shades has banned him from using it.

AI text is annoying because its optimized presentation means that it’s like a massively over-packaged product.The product itself might be okay, or it might be crap, but it’s wrapped in so much tough plastic that you have to use pliers to get into it, so its only use is to pass it on to someone else, unopened, hoping that they’ll be impressed by the package. Since they probably won’t open it, either, you’ll get credit for giving a gift.

MG is hoping no one will open it, but only because he uses it to troll and not because he intends it to be actual useful information. He doesn't want credit for a gift, he wants his superiority to reign. The problem (for him) is his use of AI is so blatantly troll-like that no one sees his 'superiority', and no one is fooled.

Your posts, along with your cabal's posts, are hopelessly stupid. Nobody gives two damnits what you pretend someone else thinks or hopes. LOL. You may be correct once in a thousand guesses. Nobody cares. It is just a yawn.

[canpakes]

Your posts, along with your cabal's posts, are hopelessly stupid. Nobody gives two damnits what you pretend someone else thinks or hopes. LOL. You may be correct once in a thousand guesses. Nobody cares. It is just a yawn.

This didn’t last long:

Addressing the contemt without making every goddamn post about the poster is also fine.

[Whiskey]

I mean that, whatever you did, to get past me you will have to have done enough actual learning that you legitimately deserve to pass. And in the same way I reckon that simply recognizing when an AI has made a good point concisely, and posting it, is a worthwhile human contribution to a discussion, not really different from tossing in an apt quotation.

This point, the point about doing what you did to pass and you passed then you deserved to pass... is getting to the core of this conversation and of AI.

We are all using laptops, desktops and smartphones to read and access this conversation. We all use autocorrect and spell check. AI has some commonality with autocorrect, right? On a basic level. If someone uses AI to clarify a point about Mormonism or another topic down in paradise, and it is not obvious, so what?

I had to study for a certification exam recently. A five hour exam. One hundred questions. Standard testing center protocols with proctors and all that stuff. I used AI to create a 27,000 line script using every question of every published practice exam, sample problem in the study material, homework problem in the online course and homework problem in the mandatory 4 day class. The script ran a practice exam that was all multiple choice. I had, in the end, about 900 questions as the script self generated more questions to correct errors that were found in the existing homework.

By the time I created the script and sample tests, I knew that crap cold. I could identify every error in past tests, past homework and the study material. I ended up creating a revised formula sheet for the certification agency to correct their published errors.

TLDR. AI as a method of studying and writing is not the same as AI as Google Answers. Not even close.