Computer grading

[bobsav21]

Does anyone know what happened to the " computer grading' system that coinhelpu was working on an while back ?

[Marvic]

Coin Slab companies decide they can make more money by doing it themselves with humans instead...

[Dutch-Tigger]

Does anyone know what happened to the " computer grading' system that coinhelpu was working on an while back ?

Greetings bobsav21, I cam across this article you may find an interesting read.
https://www.coingrading.com/compgrade1.html

[Alan H]

Interesting subject. While no doubt computer grading would be very accurate, I see one problem: Eye appeal. How would you write a program to judge that? I've had several buying experiences where I've chosen a coin that is a point or so below another because the lower graded coin (say MS-63 vs MS-62) had better eye appeal. That may come down to, say, an unsightly contact mark on Liberty's cheek on a Morgan, or an unsightly spot, or similar small thing. I think we may be stuck with humans here for the foreseeable future. Of course, the crack out game argument is also quite valid, but I'm not sure I like the inflexibility of a computer, which has been described as a machine capable of making mistakes at 10,000 times the speed of a human brain.

[Daniel]

It's on hold due to lack of funding.

Eye appeal has nothing to do with the technical grade of a coin. However, it's all light reflections that can be quantified into an AI program. AI learns very similarly to us because we created it.

[Alan H]

It's on hold due to lack of funding.

Eye appeal has nothing to do with the technical grade of a coin. However, it's all light reflections that can be quantified into an AI program. AI learns very similarly to us because we created it.

True. Technical grade is what it is. I’m not a professional grader or anywhere close enough to having the necessary skills, but I have to wonder how they might feel about that. Eye appeal matters much more at the cash register!

[Daniel]

It's on hold due to lack of funding.

Eye appeal has nothing to do with the technical grade of a coin. However, it's all light reflections that can be quantified into an AI program. AI learns very similarly to us because we created it.

True. Technical grade is what it is. I’m not a professional grader or anywhere close enough to having the necessary skills, but I have to wonder how they might feel about that. Eye appeal matters much more at the cash register!

There's plenty of higher grade coins that lack eye appeal, the technical grade is not effected by eye appeal, but collectors and dealers will pay less for a coin with less eye appeal, as you stated.

[Dutch-Tigger]

It's on hold due to lack of funding.

Eye appeal has nothing to do with the technical grade of a coin. However, it's all light reflections that can be quantified into an AI program. AI learns very similarly to us because we created it.

True. Technical grade is what it is. I’m not a professional grader or anywhere close enough to having the necessary skills, but I have to wonder how they might feel about that. Eye appeal matters much more at the cash register!

Interesting article to read if you like.
https://www.numismaticnews.net/coin-mar ... lse-values

[Alan H]

Interesting article to read if you like.
https://www.numismaticnews.net/coin-mar ... lse-values

Interesting, and pretty much in line with what I was thinking. I have chosen eye appeal over technical grade a couple of times. Usually just a point or two. One was a Morgan, and another was a Washington quarter. The thing about this is it's not quantifiable. Could the same decision making process that I went through "on the buy" push a grader to grant an additional point or two? It's not supposed to by the letter of the law, but one has to wonder. I think it does. Computer grading would take all that out of the equation. It'd also kill the re-slabbing game, so I think powerful market forces would lobby against it. We may be looking at some interesting times to come!

[Dutch-Tigger]

But still you don`t want to forget eye appeal doesn’t figure into the grade assigned to a coin.

[Dutch-Tigger]

It is going to be interesting as time goes by how much more the Eye Appeal may play into grading though .

[Daniel]

The worst thing people can do from here and into the future is underestimate AI, I've been saying it for years. If I had the funds I would have a AI Coin grading company.

[Earle42]

Coin Slab companies decide they can make more money by doing it themselves with humans instead...

Actually what the OP is asking about is a different system, an AI system, that was being developed by Daniel a couple years ago.

The coin companies initially had computer grading in the late 90s, and this was the system abandoned possibly b/c it killed off the large profits from people cracking out and re-submitting the same coins hoping to get a higher grade. When one true grade was given, a coin would only be submitted (income for the company) once instead of possibly multiple times.

As to eye appeal, that has always, to me, seemed a moot point anyway. Eye appeal is in the eye of the buyer. There are many high grade coins that I have seen slabbed that, to me, have ugly toning (not bright colors).

[pnwmakes]

As a former Aerospace quality Engineer and now finish up in two weeks on my software engineering degree. And have study AI a lot over the last year. You can teach AI not only the technical side of grading but also the eye appeal. They learn what we teach them. There is AI being used in the manufacturing industry now that can spot the smallest flaw of a product as it races down a assembly line at speeds a that look blurry to the human eye. Modern AI is so fast it probably could grade a coin in mere minutes if not faster, if it was taught how. Again it learns what we teach. AI is scary but cool as blazes.

[Earle42]

AI is scary but cool as blazes.

Very, very true on both counts.

[pnwmakes]

AI is scary but cool as blazes.

Very, very true on both counts.

The word I typed was auto changed to "blaze" but you get the point.

[Marvic]

You can teach AI not only the technical side of grading but also the eye appeal. They learn what we teach them.

Sorry, but I don't agree with your statement. A programmer's likes might not be my likes ("eye appeal"), so then we all know AI will compute the programmer's eye appeal inputs.
The AI model doesn't "like" or have a concept of "eye appeal" in the same way humans do. It attempts to generate responses based on the patterns it has learned from the data it was trained on. The model's responses are not driven by personal preferences but rather by the statistical patterns it has observed.

[pnwmakes]

You can teach AI not only the technical side of grading but also the eye appeal. They learn what we teach them.

Sorry, but I don't agree with your statement. A programmer's likes might not be my likes ("eye appeal"), so then we all know AI will compute the programmer's eye appeal inputs.
The AI model doesn't "like" or have a concept of "eye appeal" in the same way humans do. It attempts to generate responses based on the patterns it has learned from the data it was trained on. The model's responses are not driven by personal preferences but rather by the statistical patterns it has observed.

I was not talking about a individual eye appeal but a overview of eye appeal. If you taught a AI model what "eye appeal" is it could distinguish it. Now would it be the same as yours Marvic no but it could if it was taught your opinion of eye appeal.

[Daniel]

You can teach AI not only the technical side of grading but also the eye appeal. They learn what we teach them.

Sorry, but I don't agree with your statement. A programmer's likes might not be my likes ("eye appeal"), so then we all know AI will compute the programmer's eye appeal inputs.
The AI model doesn't "like" or have a concept of "eye appeal" in the same way humans do. It attempts to generate responses based on the patterns it has learned from the data it was trained on. The model's responses are not driven by personal preferences but rather by the statistical patterns it has observed.

You disagree because of a lack of vision and an understanding of AI, eye appeal is light reflections and what is pleasing might not be universal to everyone but it's near universal. Most people can agree a sunset or sunrise is eye appealing and all AI has to do is be taught that and then learn to measure the light, hues, optics, etc. It can then learn what is typically eye appealing and define it using that information. Just never underestimate what AI can or will be able to do.

[Marvic]

Thank you, Mr.D, for your view on Al's future determination regarding "eye appealing."
My understanding is AI generates what humans put into it, different personal eye appealings.
Or do I have this Ai thing all wrong? But you must remember I'm old!

[Daniel]

AI can measure eye appeal using optic measurements and the light spectrum, it can learn what is most pleasing based on learning what is already considered eye appealing to the masses. It will never be 100% but eye appeal isn't the grade and no human grader can say what eye appeal is for everyone either.