Editor’s Note: Retraction of article containing fabricated quotations
- Thread starter JournalBot
- Start date
I just got a report from this thread complaining that a user was "posting too much", and man, if that was a bannable offense a lot of you would have to find something better to do with your day.
Carry on!
"User is interacting with our community"
Upvote
37
(37
/
0)
To back up your point: a difficulty I've had with LLM review is that even when it's wrong, it usually still makes a persuasively-presented case. So it takes more effort than feedback from someone who's presenting an observation and evaluating it even as they decide how to phrase it. I know personally I've started to write parts of a review and decided against a point as I realized it's not as good a point as I first thought.
Upvote
31
(31
/
0)
Resistance
Wise, Aged Ars Veteran
It's so tempting to find an excuse to write 3 different replies to this in the space of about a minute. Edit: Wait, are you trying to get us to comment more?I just got a report from this thread complaining that a user was "posting too much", and man, if that was a bannable offense a lot of you would have to find something better to do with your day.
Carry on!
Last edited:
Upvote
12
(13
/
-1)
His excuse reads like an addict trying to explain away the addiction. He was sick…and so did the most convoluted thing possible to avoid pulling a quote from a 5 paragraph blog? Should we really believe the article is human-written and he only used AI for the easiest part of job?
Upvote
68
(68
/
0)
Snarky Robot
Ars Legatus Legionis
It cannot “help track down quotes” without creating content. Selecting the quotes you want to use IS AUTHORSHIP. If I cite a case and want a quote from it, selecting which quotation best benefits my argumen is part of my specialized skillset. That quote is part of the brief I turn in or what I say at oral argument. It is, while a quote, also a fundamental piece of MY content. I stake my credibility to what I turn in, including the portions I choose to quote.
I don’t need to pretend that marketing BS makes what some dorkass middle manager thinks of as AI comparable to what some other person thinks of as AI. I don’t need a poll. I don’t need a study. None of that matters to what I said. Common sense (and common law) matters.
If you are writing something that contains quotes, then the quotes you select ARE PART OF WHAT YOU WROTE. If you allow a computer to select those quotes, you have just used generative slop machines to dictate your content. The machine wrote part of what you are putting your name on. If there’s a policy against doing that, you’ve violated the rule.
The reason you can put together a book of recipes and have it be copyright protected even though a single recipe cannot be is because selection and arrangement are, in fact, a form of authorship.
It’s not wholesale letting the slop machine write the article, no. It’s the equivalent of, say, using slop machines to generate the opening lines, or the title and blurb. But Ars policy would require disclosing any of that because it’s actually part of the content you turn in to them and hold out as something you alone authored.
Last edited:
Upvote
62
(63
/
-1)
Ahem, the report is clearly "User is interacting with our community excessively"
Upvote
11
(11
/
0)
anguisette
Wise, Aged Ars Veteran
since this is Ars, i really hope someone will do a statistical analysis of post counts on this thread so we can find out who was posting too much... wait, hang on...
okay, i've been informed that this actually means "posting too many things i disagree with". ignore me, i am not au fait with today's youth slang.
Upvote
36
(37
/
-1)
Eufy had a scandal where they exposed video and then lied about it. Ars covered it. A few weeks later they posted a dealmaster talking about how great Eufy was. Same thing with Ring a few weeks before that. They reduced the dealmaster stuff which was nice, but they didn't do anything to the post after the comments. It's only 3 pages of comments and mine are in there. I also emailed Lee, but it was pretty much my comment. I wanted to make sure he saw it.
https://meincmagazine.com/gadgets/202...lock-is-now-on-sale/?comments-page=1#comments
On another note, I'm really tempted to put a canary phrase in my emails. "No AI was used in creating this." And then when I do use AI, I'll delete it. Would have to build the trust and be good about it, but I don't use AI that much for documents I sign my name to and I'm tempted to use it less.
Upvote
22
(22
/
0)
Now that the conversation has shifted to who's using which bathrooms and who's posting too much, it's probably time to close this thread.
Upvote
-7
(12
/
-19)
anonymouschicken
Ars Scholae Palatinae
I've been feeling some AI-related anxiety recently. I'm in the process of buying a house (first-time homebuyer!) so I think the prospect of committing to a mortgage has awakened some "holy shit what if these fucking chatbots actually do steal my job" energy that I'd been avoiding so far.
Anyway, last night I asked the free version of chatGPT to list the numbers from 0-100. It did this correctly. Then I asked it to limit the list only to multiples of 3. It did this correctly. Then I asked it to filter based on numbers that rhyme with "knee." Arguably it did this correctly, including all the numbers ending in "3," but omitting numbers like 30 and 60 that I personally feel should have qualified. Then I asked it to expand the list to include numbers rhyming with "won" and "heaven," expecting that numbers like 21 and 27 would be added to the list. This was the tipping point, the list it returned was nonsense, including many numbers that weren't multiples of 3 as well as omitting many numbers that should have been included based on the earlier rules. This wasn't improved when I used subsequent prompts to point out its mistakes and clarify my instructions.
I came away from the interaction feeling a lot less anxious. These bots don't think, they don't reason, they still make blindingly obvious errors. And the worst part is they still present their mistakes with complete confidence, even after being corrected. "You're right, looks like I messed up there, here's a corrected [but actually still completely wrong] response!"
Don't rely on these bots for anything important. Always verify whatever they give you.
Anyway, last night I asked the free version of chatGPT to list the numbers from 0-100. It did this correctly. Then I asked it to limit the list only to multiples of 3. It did this correctly. Then I asked it to filter based on numbers that rhyme with "knee." Arguably it did this correctly, including all the numbers ending in "3," but omitting numbers like 30 and 60 that I personally feel should have qualified. Then I asked it to expand the list to include numbers rhyming with "won" and "heaven," expecting that numbers like 21 and 27 would be added to the list. This was the tipping point, the list it returned was nonsense, including many numbers that weren't multiples of 3 as well as omitting many numbers that should have been included based on the earlier rules. This wasn't improved when I used subsequent prompts to point out its mistakes and clarify my instructions.
I came away from the interaction feeling a lot less anxious. These bots don't think, they don't reason, they still make blindingly obvious errors. And the worst part is they still present their mistakes with complete confidence, even after being corrected. "You're right, looks like I messed up there, here's a corrected [but actually still completely wrong] response!"
Don't rely on these bots for anything important. Always verify whatever they give you.
Upvote
29
(29
/
0)
anguisette
Wise, Aged Ars Veteran
"ChatGPT, call my interviewee at +15551234567, ask them some insightful questions, and provide a summary of the best quotes i can use to write my story, which was definitely not generated by AI" -- typical Vanity Fair reporter using a mere tool to automate the boring mechanics of their job.
Upvote
25
(27
/
-2)
I always figured this was part of what's behind Ars' comment quoting policy (and yeah, I bent it a bit right here, by not including the entire quote myself).
Quoting has two burdens: 1) the quote itself and how you present it is your own work. 2) the citation links to the work it came from, so others can understand the context.
If you mess with either of these, misrepresenting something as your own work, or misrepresenting what you're referring to, those are both bad.
AI makes it difficult to validate BOTH burdens, as it can change what you present to be something you'd never present yourself, AND it can make up what other people said while sounding very convincing. The end result is that AI cannot be the final reviewer. Ever. Which often means that depending on an AI can result in more work than just doing everything yourself.
Personally, I find AI to be a great writing prompt; I get it to generate something, read through it, say "that's not it at all!", then possibly get it to do a few more attempts, and by that point, I know what it is I meant to say, and can write it up in my own voice with proper references.
Letting AI do the talking is always tempting, but it's a temptation that always has to be either avoided or cited (yeah, you can cite an AI as the content source. It lets everyone else discount it as meaningless, OR see the value in it, if there is any).
Upvote
11
(12
/
-1)
Just doing my part to ensure CN's feed into the data maw results in a future AI-fed world filled with the Tao of graylshaped...I just got a report from this thread complaining that a user was "posting too much", and man, if that was a bannable offense a lot of you would have to find something better to do with your day.
Carry on!
Upvote
15
(15
/
0)
anguisette
Wise, Aged Ars Veteran
why? was the issue resolved? somehow i missed that.
i'm okay with getting ejected from the thread for talking about bathrooms, but it's pretty absurd to let a couple of trolls close an entire thread, don't you think?
Upvote
32
(34
/
-2)
Or, people can just talk about whatever while they're patient, because if we locked front page discussions for going off topic we'd have to close like 90% of them.
All I need is for people to not be abusive or personal.
Upvote
70
(70
/
0)
If you ask a search engine to help you track down a quote, you know that it's going to show you what it finds. If you ask AI software to act as a search engine and help you track down a quote, you don't know for sure that it's going to show you what it finds or if it's going to try to please you by making up something that sounds like it could be something it finds.
Either way I don't have an issue with using a tool to track down a quote, but I do think using AI to do it should come with the expectation that you are going to do so much work on your part to confirm that the quote is in fact real that by the time you are done you have taken 100% ownership of the content you are putting under your name. And at that point I think it's passed beyond the intent of the policy, if not the letter and shouldn't be an issue. I'm saying this as someone who is very much an AI skeptic and wouldn't use one for paid work at all.
Upvote
26
(26
/
0)
But can we stay away from discussing who is spending too much time in the bathroom?
Upvote
2
(6
/
-4)
Thank you for introducing this topic. I look forward to the discussion it will engender.
Upvote
5
(7
/
-2)
Can't really trust search engines anymore as they're getting increasingly tied with AI...
Upvote
26
(26
/
0)
anguisette
Wise, Aged Ars Veteran
you're not wrong, but at the same time, you're a little bit wrong. LLMs are awful at questions like "generate a list of numbers from 1-100 which rhyme with heaven", because 1) they're bad at maths problems, and 2) despite being language models, they're terrible at processing language, hence my earlier anecdote about having ChatGPT review copy. they don't deal in words, or text, they deal in tokens, and it's pretty difficult to take a series of tokens and identify things like rhymes or spelling errors when the LLM doesn't "see" the actual text.
basically, if your job involves generating lists of numbers from 1 to 100 that rhyme with "heaven", you probably don't need to worry. but if your job involves writing text...? you can give a list of basic facts to an LLM, and ask it to generate a story reporting on those facts, and it can do that, and it can do it well enough--most of the time--that it could even pass for journalism. it's not actual journalism, because it lacks that spark that good journalists have to craft engaging prose, but it's good enough that at least one local newspaper, mentioned in this thread, uses that process to write all of their stories.
does it sometimes make things up? yes. does it sometimes misunderstand the context and get things backwards? yes. should anyone trust that sort of "reporting"? no. will your boss care when he's laying you off because an LLM has taken your job? no.
that's the problem: LLMs do not have to be good enough to replace humans. they only have to be good enough that managers can replace humans with LLMs and save money.
Upvote
11
(16
/
-5)
I just realized that AI users are exactly like the lazy scientist who built a robot from that one Scary Door bit on Futurama.
Upvote
15
(16
/
-1)
Snarky Robot
Ars Legatus Legionis
The funny thing about having to debate some of this nonsense is that some people might genuinely have an easier time understanding this if AI stood for Ars Intern.
If Benj had sent an actual, factual human intern to go read the blog and pull quotes, that person would have helped write the article. That’s obvious. Benj would have a responsibility to check before publishing. He, the senior reporter, would be responsible for the Ars Intern making up quotes and allowing them into the article.
Similarly, if Ars created a policy where if anything the intern did resulted in content that was part of the story, that had to be disclosed (amyou’d put more than one set of eyes on it just for legal review), and if, as he did here, Benj assigned the quote selection to the Ars Intern, he would have violated that policy.
The same errors happened here with AI being a slop not instead of a human.
Upvote
51
(51
/
0)
At least with the ones I use if I ask them to track down a quote they are going to present me with a list of links to pages that already exist and which match to some degree the criteria I asked them to search for. Now, those pages may be infested with AI garbage but if I'm only looking at search results and not the clearly identified text generated by the search engine's own integrated AI, the problem you cite isn't an issue.
Upvote
8
(8
/
0)
Snarky Robot
Ars Legatus Legionis
There it is folks: tacit permission to make this Pony Thread 2: Derpy Boogaloo.
Upvote
28
(29
/
-1)
Confusingly, Copilot has now gained a Search mode which does focus on giving you the citations. At the moment that is separate to Bing Search and, unfortunately, it is vastly better than keyword search in the case where you would otherwise have semantic difficulties, like searching for stuff about the band "The The."
But it's quite possible that once Microsoft have kicked that around a bit they'll put it behind Bing by default, and then you will indeed be using that engine whenever you just use Bing.
(The only thing that can save us here is that no-one uses Bing.)
Upvote
2
(4
/
-2)
Pft, I'd have been banned years ago.I just got a report from this thread complaining that a user was "posting too much", and man, if that was a bannable offense a lot of you would have to find something better to do with your day.
Carry on!
Upvote
7
(8
/
-1)
The only time I even think about Bing is when I remember reading when it was launched that they intended it to become a verb the same way "Google" had. I immediately saw the problem with that is that for any native English speaker the natural past tense of "Google" is "Googled" but the past tense of "Bing" is "Bung." Picking that name was another classic Ballmerian bungle.
EDIT: Yes, I know it's more like the past participle than the past tense ("I have rung" vs "I rang") but it's still where the mind goes.
Last edited:
Upvote
16
(16
/
0)
Well by that example even the correct past tense would be Bang, which still leaves you in similarly amusing territory.
Upvote
11
(11
/
0)
Is it possible to filter articles written (heh) by a specific author? At this point, nothing coming from Benj Edwards can be trusted.
Upvote
7
(11
/
-4)
I'm not going to argue against this.
But for every tool that can be valuable if used correctly, there are two questions: how likely is this tool to be used incorrectly, and what are the consequences if it is?
A car is a tool that can be valuable if used correctly, but the consequences if it's used incorrectly are such that most places require a license to legally drive one on public roads.
Generative AI is obviously not a tool that is necessary to doing journalism, as the profession is much older than the tool.
So: how likely is this tool to be used incorrectly? Well, a person whose subject matter expertise is AI managed to screw it up, so I'm going to say "almost certain."
And what are the consequences if they do? Well, look at the past 35 pages of comments for what something like this does to a publication's reputation.
So that's a near-certain chance of misuse, a catastrophic consequence if misused, and the profession of journalism got by just fine for decades or centuries without it.
I don't think it really matters how "valuable if used correctly" a tool is, in that situation. The risk of misuse you're taking on if you allow it is too high.
Upvote
32
(32
/
0)
"AI" users overlap strongly with the target demographic for digital Ron Popeil products.
Not all of course. Anyone taking exception to this characterization is, of course, different and special.
Upvote
8
(8
/
0)
Most people don't have 100" TVs nor are especially fussy about quality. Also, most ways you'd watch an actual DVD on modern equipment will have some sort of upscaling happening. (even on my low end, decade old TV, a DVD on a player that's about twice as old looks comparable to HD cable from normal viewing distances.) Though "DVD" is sometimes used generically too.
Upvote
1
(7
/
-6)
I thought there already was an attempt at a second Pony Thread, back in the end days of the old forum when the number of pages of replies was causing Moonshark errors.
So this would be Pony Thread III: Hay Hard With a Vengeance.
Upvote
13
(13
/
0)
anguisette
Wise, Aged Ars Veteran
this is a real problem, and not just with generative AI. it's like, you can build a tactical nuclear weapon to take out the skunk living in your back yard, but once you do that, aren't you going to be tempted to deploy the next one on your neighbour's yappy dog? and there aren't even any consequences to doing that. some weirdos are like "i don't think everyone should own tactical nuclear weapons", but let's be real, the market has spoken, and nukes are the next serious investment opportunity.
at this point, it's simply unrealistic to oppose private ownership of tactical nuclear weapons. the conversation has moved on, and there's no point re-litigating ancient history. we might have to endure a short period of mass destruction and people dying of horrifyingly painful radiation poisoning, but once we get past that initial hurdle, you'll see that private ownership of tactical nuclear weapons just makes everyone so much nicer, and more polite. who's going to wolf-whistle you on the street when you might have a nuke in your pocket?
anyway, as a practicing journalist, i sometimes use tactical nuclear weapons to prepare stories (i NEVER use them to generate prose) and no one has ever complained that i'm doing a bad job. well, except my old boss. but i have a new boss now, and we get on great.
Upvote
38
(39
/
-1)
Upvote
6
(6
/
0)
- Status