Using Artificial Intelligence (AI) in Coin Collecting

That sounds like a fun little side project for you!

 

Log in or Sign up to hide this ad.

Like he doesn't have enough to do. Feed the Dogs, Cats, Hog, do homework, do dishes, sweep the floor, paint the fence.....

 

I occasionally use Google image search for strange numismatic items and the photo app I use on my phone has an AI button that has helped my terrible photos look better. Other than that, I haven't really messed with anything else. I could see the potential benefits of using it though and I definitely think AI will increasingly become heavily integrated into our lives.

 

I have commented on this before but will have my say. there are 2 major problems with current AI, (This is also stated by 3 of the creators of current AI). 1 Everything they do or use is derivative. this is to say they have no original thinking; they only respond to data, period. 2 they have been programmed to discount all human involvement as flawed but to accept all computer/AI information as accurate. This means if even 1 AI program finds and acccepts flawed information all AI programs now acccept the incorrect information as fact. The creators admit they were forced to rush a product to market and all 3 admit making this mistake. It will only be when an AI program is created or reprogrammed to be a true sceptic that AI will be able to advance.
I have also said on other sites that there is only 1 way I see AI gaining traction in the hobby. They could set up a CAC type stickering program to offer either more advanced opinions on coins in existng slabs (IE, think a grade of AU-54.2), or to recognize varieties, die marriages etc. (Messydesk in a box). James

 

I'm not sure what it does exactly but it balances all of the settings to make the photo better. All of the lights in my house are either too blue or yellow when I take photos and it evens that out.

 

Wouldn't the analysis of a coin only be as accurate as the photo taken (by a human) is of high quality? There are a lot of bad photographers out there (just look at ebay), which would make it impossible for AI to accurately analyze a coin, no?

 

@Eric Babula . My biggest concern with the AI when it comes to grading is that something a person might see and take into account; something like a weak strike, mild cleaning, plnachet flaw etc. ; could easily become part of the "standard" for a particular grade. Nex thing anyone knows the computers are downgrading coins because they don't have a small gouge near the date or think all coins of a specific grade must have a certain toning. I think human riview is going to be a constant thing; the next problem is convincing the AI program that the human is right. James

 

Some of this AI stuff is really impressive -- and scary.

 

That doesn't look anything like the 1836 to me, not even close. The first one however looks like it could be the model used for the new dime.

 

The AI picture looks prettier (to me) than what I would imagine from the coin. YMMV.

 

Google's AI seems to have hallucinations about some coins. I just acquired a number of Morgans marketed as "XF," and they are indeed nice. One 1879-O looks extra nice (CoinKnow rated it a 60, and I agree) so I asked Google AI about the value. Among other things, it said the following:

"Variety Check
Look closely at the reverse. There is a rare variety called the 1879-O "Large O" (where the mint mark is significantly bigger). If you have a confirmed "Large O" in XF condition, the value can jump to $250–$400+.

To identify the 1879-O Large O vs. Small O, look closely at the mint mark on the reverse of the coin, located above the letters "D" and "O" in the word "DOLLAR".

The Difference

• Small O (Common): This is the standard mint mark used for the majority of the 2.8 million coins struck in New Orleans that year. It is a relatively small, neat oval.
• Large O (Rare): This is a specific variety (often cataloged as VAM-26) where a larger "O" punch was used. It appears thicker and fills more of the space between the eagle's tail feathers and the wreath."

I got a bit excited, because the "O" definitely looks larger on this one than the O on some other dates I have. I tried to find comparison pictures or any other info on this and found nothing anywhere. I looked and looked! Today, I went back to Google's AI again, and now it says this:

"
While the term "Large O" is often used by collectors to describe a standard mint mark, the 1879-O Morgan silver dollar does not have a officially recognized "Large O" vs. "Small O" variety in the same way the 1879-CC (Carson City) does.
However, there is a very rare and valuable variety called the 1879-O/S (O over S), where an "O" was punched over an "S" mint mark. If you are seeing something unusual about the mint mark, it may be one of these:"

So that kind of let the air out of my balloon. Has anybody ever heard of this "large O/small O" mint mark business? Or was this an AI hallucination?

 

I came across a version of chatgpt called numi. I've started to use it a little.
I don't yet know that I'd trust it own its own but with foreign coins that I'm not sure on it can identify them and then i can ask it questions about the coin. It at least points me in the right direction.
Its also able to grade coins but I find its lacking and sometimes needs walked though the process. I would not trust its grades on its own but on things I'm not familiar with i can get a quick starting point at least and it will give guidelines for the grade.