AI Chip Pricing? (12 Viewers)

[ReallyGoodUsername]

Has anyone ever used AI to price a chip?

A seller just sent a screenshot of AI pricing they got. It basically said the range was $100-$150 and $140 is what they should hold out for.

I checked Worthpoint.com which is a paid website that tracks past eBay and other auction/sale sites. It’s by no means perfect but is useful for uncommon and rare chips like the one being sold. It showed a handful of list prices (sale price may be lower if a best offer was accepted) which were in the $50-$70 range.

It’s not a priority chip for me so I just passed along the information but it made me curious if anyone knew how accurate pricing models actually are for this type of item? Is it incorrectly grouping other similar chips together or is it playing 4d chess and the past sales are no longer inline with the current market?

 

[GamingWithChips]

I can't speak for all models, but my quick assumption is to not trust AI for things, let alone something this niche. Maybe in the coming years AI will actually do a good job of scraping the internet and analyzing prices to produce an accurate value, but they are still way too easily influenced by things they shouldn't be. (ex someone posted a random fake 'fact' on reddit, then searched for the fact and an AI said the fact that he made up was 'true'...it then sourced his post...lol)

FWIW I had to use an AI scan to value and mail chips from Canada to the US and it arbitrarily gave a price that had zero resemblance to the actual value of the chips. Just a complete guess from what I gathered.

 

[upNdown]

I just recently started selling some trinkets on eBay, for the first time in a while. I was surprised to see that as I put together the listing, eBay’s AI told me what my bin price should be.
That also seemed to be inflated bullshit.

 

[JeepologyOffroad]

I would never ever trust something like that. Especially not for such an inside baseball market like chipping.

 

[Zmasterben]

Oh God, "AI told me what its worth" is going to become the new "I know what I have" across all hobbies, isn't it?

 

[DrStrange]

The AI is only as good as the data it trains with. Feed the AI a few "chip guides", toss in some old CCA magazines and I expect to get well overpriced results. Same thing with single pricing for rack quantities.

Chipping is an "insiders" game. The AI doesn't have the right information to do an acceptable job -=- DrStrange

 

[Cratty]

Compute x Context x Model Quality = Result Quality.

But yeah, if compute and model quality keep scaling as they are… I assume it wont be long before the models can do better on average than most of us can, because they can then compensate for lack of context/data being fed into the prompt. For sure not there yet though.

And on that note, I’m now sad again as I start to think about the future job market. Time to go to bed.

 

[GamingWithChips]

And on that note, I’m now sad again as I start to think about the future job market. Time to go to bed.

I mean, in 3 years it'll be 2029, just enough time for AI to get better and perfect timing for another great depression, we're due.

 

[FDLmold]

I would equate somebody using inflated AI pricing to somebody using inflated "by the book" pricing. I've encountered the latter, but never the former. When interacting with either, however, that seller would immediately drop to at least D-tier, and probably F-tier. That being said, chip prices can move unexpectedly, but almost any $50+ casino chip is low supply low demand, and getting a feel for current pricing frequently lacks data.

 

[AlbinoDragon]

I frequently mark up my copy of the Chip Rack with recently sold prices I see on eBay for chips I either have or want, mostly so I can track what the "current market value" is for some of these.

I'm surprised at how close 'book value' is at times. I'm shocked at how far off that is the rest of the time... both high and low.

At some level, they are worth what someone is willing to pay at that moment. There's been sufficient chatter about WindWalker in the last week or so, I don't feel too out of place invoking a chaos-agent to disrupt what one might think of as "normal".



Closer to the question at hand, but as @GamingWithChips , @JeepologyOffroad and @DrStrange mentioned above, AI will only ever be as good as the model it's trained on. I can see a case where one gives you a radically different answer than another, simply because it wasn't pointed at the correct training set.

A couple weeks ago while I was at the observatory, there was a crew that arrived to being the installation of a brand-new instrument for the telescope. Similar ones have been built and are in use elsewhere, but this one is truly "the latest and greatest" for what it will do. The crew included a couple eager and quite sharp doctoral students and while we were unpacking crates of the bits and parts that make the magic work, one of the students asked the principal investigator (P.I.) for the project, "do you think A.I. will revolutionize what is done with this?" Without a beat, the P.I. replied "with what training set? Nobody has done this before, and whatever A.I. we'd use would be the one we train with our brand new data."


If you are looking at a small enough training set (or just a small enough user group looking for that answer), A.I. will miss the answer for some time still. It's not an all-knowing font of knowledge, it's just pretty good at the most common inquiries.