Case study in AI: my personal experience, and how it affects listings
Summary: countessalmirena starts a discussion on their personal experience with AI and its impact on listings. twyngwyn questions the involvement of a platform named 'Khoros,' while scottydbm agrees with skepticism towards AI, emphasizing its artificial nature. sandypebbles shares a warning from cyber spy chiefs about AI, suggesting an emerging threat. The discussion reflects concern over AI's role and potential negative impact.
Apologies in advance for the length of this post.
This is a worm stew of hallucination dressed as certainty - by which I mean AI will lie to your face even when challenged.
It's a cautionary tale about using AI to gain and to disseminate helpful information - so it's relevant to buyers and sellers on eBay, and to anyone posting on these boards to give advice to eBayers seeking help.
It's also relevant to sellers relying on AI-generated information for their listings. Pshaw. Pfooey.
The Situation
I happened to be looking for some information regarding where and when a specific CD was recorded. It doesn't really matter which recording, but for the purposes of illustrating the point, it's a recording of Gounod mélodies by one of my favourite baritones.
Luring AI into the matter
Spoiler - can be skipped unless you want to read the actual details
Then I allowed AI to enter the picture, as a test. It blandly informed me that the recordings took place between such and such months in 1994, and were in a particular studio in Lyons. Pressed to give details of its sources, it provided them. I read each one in detail, and nowhere was the information confirmed. The AI response, when I added that I was unable to verify the details and stated that I thought it might be hallucinating, was a firm NO, that it wasn't hallucinating, that the information was verifiable. More links.
I read each one of those, and informed AI that I owned the disc in question, that nowhere were months and studio included in the liner notes or booklet and disc and artwork, and that furthermore, the new links did not include that information either. AI continued to affirm its "truthfulness" and slavishly praised my scepticism. (Ha.) More links. I read those too. Again - nothing at all to confirm the details, and at least three of the linked materials weren't even relevant.
Short summary: AI gave me false information. I challenged it. It maintained that the information was correct. Multiple links were supplied. None of the links actually contained the information that AI claimed they had. I continued challenging.
I then added my coup de grâce:
AI confesses to the crime
At that point, AI grovelled (a feature that I truly loathe) and admitted it did not have verifiable information and it was wrong and it apologised - which is just another piece of silliness. It hallucinated faux information by drawing connections from disparate sources; it didn't act through human emotion. It's nonensical to say that it is capable of actually apologising - but that's by the way.
The main point is... AI will indeed hallucinate information that looks sufficiently credible to pass muster unless you're an expert, or feeling grumpy and disinclined to trust AI, or just don't trust AI (no need for grumpiness), or have done some research yourself without AI sticking its nose into the whole thing.
Relevance with regard to answering questions on these boards
I've already mentioned my suspicions about AI being used on the new iteration of these boards. I hope I am wrong about it. In case I'm not, I am hammering home the danger of relying on AI for accurate information.
It may look like a duck, it may quack like a duck ... but it might be an AI rubber duck fit only for throwing into the bathwater. AI is not reliable. It's not. The worst thing is if the AI summary or answer aligns with what the questioner already thinks (roughly, along those lines - that sort of thing), because the overwhelming temptation is to say to oneself "Well, that's all right. That's what I sort-of thought."
Even challenging AI and being sceptical is not going to result in a more accurate response. One has to push back - and push back hard.
So... Answers given about an eBay situation will be based on inaccurate information - or irrelevant information. In some cases, AI seems to pull in eBay information from all over the globe, or from previous eBay iterations, or disparate information that links unrelated bits and pieces into a misleading whole.
When humans post on the boards, the user doesn't feel as though the posts are holy grail. They can be challenged, discussed, disagreed with, agreed with in part, confirmed, denied... It's humans interacting with each other.
As much as possible, we'll include links to the relevant eBay Help pages, but we tend to answer as well, even if it's a short or summary sort of answer with the added exhorting of "Please read the link".
But when it's an AI-generated answer, it's presented as being more authoritative. It's a known phenomenon.
That is in itself misleading. AI is in no way more authoritative. It's a mistake to think of it in that way. It would be a grave mistake if eBay goes down the AI route for the boards.
Relevance for sellers
I'm a buyer who collects things. No need for me to go into details because the relevant thing is not what I collect, but that I, like many eBayers, collect things specific to my interests. Almost any collector wants accurate details about specific items relevant to their passions. They don't want generalised and inaccurate slop. It doesn't help buyer-collectors. We need the details and we need those details to be correct.
Collector's Diatribe (skip if you don't want to read)
It's possible that there are buyers who don't care about the details and will choose by what's there. But when - for instance - looking for a particular edition of something, or a specific variety, or what-have-you - when the details matter, AI slop is no substitute for accurate information.
Now that's off my chest.
sandypebbles
·1 month ago‘Act now’: Cyber spy chiefs issue warning on AI threat
scottydbm
·1 month ago · Editedcountessalmirena, You should be Grand Duchess countessalmirena, Yes, everything you have told us about AI is TOTALLY 100% Spot On. AI CANNOT BE TRUSTED, Why because it is Artificial, ..... yes, NOT REAL !!! I Rest my Case, Your Honour.
twyngwyn
·1 month agoAnd now, it’s running the ‘boards’. Khoros where are you?
scottydbm
·1 month agoKhoros Has Been AI'd.
domino-710
·1 month agoFirstly - in each and every search - one can scroll down past the AI - a given.
Seems there is a tad of no understanding of AI.
AI is the engine - Data is the gas.
In this ' case ' - you have asked AI for information - information your baritone himself cannot even confirm - from memory - nor can the studio in which this was held - ie - there is no DATA - so why go to AI and argue.
Just my opinion - any discerning collector of anything would go further than AI.
Even having been around since the 50's - from what you are expecting - AI is in it's infancy today.