How does Ai work for counterfeit detection?

[bashort]

So the way that I think AI is supposed to work is for example AI will analize all available data to predict what the weather will be like next week, then when next week comes it will compare its predictions to what the weather actually was and make correction to the algorithm and do it again and again getting smarter the more cycles it does.

So then we have AI doing similar to find counterfeits, it takes all of the available data and makes a prediction about if a item is counterfeit or not, the question I have is are the predictions that AI came up with just accepted 100% and not verified, if that is the case how could AI "learn" anything if every prediction is considered correct.

[coastaltechsolution]

Who knows what AI engine and API's they use?

How are they training Skynet?  😁

[evelyb30]

---

@coastaltechsolution wrote:

Who knows what AI engine and API's they use?

How are they training Skynet?  😁

---

And where's the Terminator when we need him?

[coastaltechsolution]

He is being reprogrammed by Conner to come back to 2024 and save eBay 😂

[retro_entertainment_collectibles]

There are many different types of AI even among adaptive AI.  Inputs would be rather wide ranged depending on products, pricing, images, colors, printing, etc history, like products, nation of origin and much much more.  Counterfeit Coach purse quite different than say counterfeit thermal mugs. This is a problem as manufacturers are not too keen on the idea's of sharing engineering/manufacturing data with the points of sale.  For example, a thermal mug has "nn" diameter, "nn" thickness, polished aluminum versus spun for example, etc etc. and AI can extrapolate that based on comparative model/scaling.  At the end of the day alot comes down to math, "stich size", "double stitch", color shifts on and on.

Instead, they've hired external companies that spider websites applying their AI models against listings which DO have the specifics in aiding identify counterfeits.  This where I state there is a clear conflict of interest in all this Shop Safe stuff.  These external contractors are in the business of locating and reporting counterfeits, manufactures, point of sale, legal entities etc.

Obviously if eBay or other operations had the data they could better mitigate counterfeits and these contractors be out of business.  Those contractors reporting to government committee are in a clear conflict of interest because if points of sale had the data the contractors enjoy odds are those contractors are out of the business of finding counterfeits.  So how might we expect they testify in front of committee?  They've self interests and that's a conflict of interest IMHO.

[ckimodog]

@bashort 

That's a great question and you do have to wonder.  Perhaps others can chime in, but in my opinion a hands on view will trump a computer everytime.  JMO.

Have a great evening and GO CHIEFS!

[deltilogical]

it will be used to create counterfeits, not stop them.  The videosynthesizers are being optimized for political warfare. 

[retro_entertainment_collectibles]

---

@deltilogical wrote:

it will be used to create counterfeits, not stop them.  The videosynthesizers are being optimized for political warfare. 

---

Indeed so but will it end there?

What about when AI is applied lets say to copy then manufacturer of hard goods via automation?

Its truly mind-boggling and as a software engineer myself not that I am modern AI savvy this is a best laid plans of mice and men deal IMHO.  It's nice to think that this type stuff be controlled but as we are already seeing bad actors can use it for maligned purposes and this is just the start.  If there is one thing mankind has shown his inventions are nearly always used for both good and evil.

[albertabrightalberta]

@bashort 

I've seen AI work very well in some cases, but not necessarily in detecting counterfeit items. 

The way it's supposed to work doesn't seem to be the way it is working.

As for counterfeit items, I can only speak for the category with which I'm familiar, experienced and in which I've seen so many mistakes the I'm estimating their record to be close to 50-50. (Sheesh, I could do 50-50 with my eyes shut and just pointing a finger!) The category I refer to is handbags and accessories.

The way AI is supposed to work is that first, they collect a whole bunch of data, enough of a sampling to make judgements, supposedly knowing which is "correct" (i.e., authentic) and which is wrong. As more data is collected and as they are notified of errors, they are expected to update their algorithms so for future authentications, their system will recognize those items  that are fake and will no longer make that mistake again. 

However when E*****y's clients and/or E*****y are notified of mistakes, either through messages sent through listings or as NAD disputes after the fact, it doesn't appear that the information is being input and as such, corrections aren't being made. 

Buyers of fakes who've notified sellers who use this AI software HAVE received refunds but those companies/sellers have disclosed to the buyers that they have COAs and that the items were found to be authentic! 

Personally, I wouldn't blindly trust buying a handbag or accessory that comes with a COA from an AI authenticator. I'd want a set (or 2) of eyes of HUMANS who know the brand, know the items and know what to look for to determine authenticity.