I think you're looking at it a bit wrong. They have laser scanners that can 3D map a coin out down to the micron. With that you can just train a computer to detect any moved metal. Then an actual grader can sit at a computer and look at the map of the coin, see all the statistics of how strong the lustre is, how many cubic millimetres of metal has been displaced by scratches, how sharp is the strike. And after a computer lays down the objective truth for the grader to see, they can decide where that data deserves to be on the sheldon scale. I think it'd significantly speed up the process and then you'd not have to rely on the human perception of reality which we know is flawed.

Christian, thank you for your information you share with us. I’ve been a collector of 50 years, but hadn’t done anything until 2 years ago. Seeing your videos and one other gentleman are very informative and not loud and noisy so thanks

I agree with TT"s assessment. The real litmus test for true "AI" in grading would have to compete with human graders and outgrade them on a consistent basis. Only then would you have to sell that product to the industry, get most everyone professionally grading coins fired, recycle all the coins that aren't graded by "AI" even if it did catch on, and foist a new complication onto every collector's collection, that doesn't sound fun that sounds like a mess. it's going to be a torpedo in the hull of the hobby. If you don't know whether your coin is beautiful or not you probably shouldn't be collecting coins. Many collectors rightly don't care about exact grade. It's subjective anyway, it's expensive, even impractical for people collecting lower value coins even mid-mint common-date Morgan dollars, it's not worth it economically. Does that mean that collecting those coins isn't worth it? Of course not! But to pass the litmus test, good luck with that. Every coin has a different strike, if we're talking microns here, then the variance between strikes across the coins on all surfaces is going to be immense, and yet it's imperceptible to the human eye, and thank God it is, because it doesn't matter to the collector or the collection. What problem are we trying to solve here anyway? That some people gripe about grades? smh Heck that's part of the fun. Solving the snowflakes' little problem to dehumanize/ruin the hobby. Beauty is in the human experience not coinbots telling us what our coins are. Looking at coins is subjective and reliant upon human perception. Arguments for AI being made around here assume this perfectly refined technology, AI's full potential, and pardon me but they're dreaming about centaurs. Centaurs that will eat the children of the hobby. I'm not impressed and I'm canceling AI on this one.

If AI is used instead of people than the service should be cheaper. But we all know they won't do that.

Every time I refresh my page I have to re- subscribe

Genius use of Cunningham’s law

Interesting stuff!

This is a thoughtful topic Christian. AI is ultimately based on data, and what data is necessary to accomplish task. Ones and zeros do not provide nuance, but can give approximations, not certitudes. Machines are never 100%, but give fractional observations. Additionally, machines do not use money. What is required in judging is a piece of the game. I believe the risk of being wrong makes humans better at human tasks such as coin evaluation. And think about this, AI will not be able to put two coins at the same exact grade, instead the machine will always give a spectrum where nothing can be truly matched. My two cents…

Great to address this topic ... I think your answer may be right for next 5 years but longer term I would not be surprised that they incorporate AI into the process. Sort of like a pilot flying an airplane.

Hi there you did a video for 2003 Alabama quarter with one variety will I found another one on both it was a double nose

I can't imagine AI coin grading being any more inconsistent than current NGC & PCGS grading.

AI (or ‘machine learning’, which is what we’re actually talking about here) could be used quite quickly should an organisation with access to the data (e.g. PCGS or NGC) choose to train it against agreed grades assigned by a competent human. The Controls parameters (e.g. light level used, distance from the coin for the examination and magnification level - there will be others) would need to be universally agreed but once those were in place it would just be a decision on when the sum of the inputted data generated acceptable outputs (i.e. an accurate grade). Good subject 👍 and you make a good point about older and ancient coins - just too much variety and subjectivity for machine learning right now……

The grade of a coin is a compilation of various independently evaluated factors. I'm not sure this is something AI - which is really just an adaptive algorithm tied to pattern recognition - can get right enough to rely on 100% or even 99.9%

Hey man, i have a really unique set of trinary bills.. not graded or anything, but are in un-circulated condition. 3 consecutive bills, 2 of which are trinary. Much rarer than simlly one trinary. Can we talk by chance? Im looking to sell them

Ai would be really terrible in coin grading and would takes drvades to learn. The first probelm being grading a coin from different magnifcation levels. 5x 20x 20x 25x. Each magnification makes a coins grade different. Ai i just dont see bring able to manage all the intricacies

I feel AI would benefit because there would be more consistency in grading

I have to correct you these companies are not grading 10,000 coins a day. Kmon man. You can do the math on pcgs and ngc and their company on the stock market. These companies on most average a day arent doing 10k coins a day....

Like calling strikes and balls in baseball.