AI just broke through a major bottleneck in coin analysis

I disagree (surprise!).

Yeah, chess is a discrete problem, with a finite set of solutions. But that set is really really REALLY huge. The number of possible games is much, much larger than the number of atoms in the observable Universe. There is no hope of exhaustively, mechanically enumerating them.

Depending on your perspective, the number of possible distinct measures combining to determine a coin's grade might be infinite. But the number of distinguishable combinations is not. It's small enough to fit into a well-trained human mind, just as the meaningful chess positions do.

Where we likely differ: I'm not convinced that there's some immaterial component of the human mind that computers will never emulate. I suspect that either there's no such component, or there's nothing stopping computers from emulating it along with everything else. Or maybe even gaining it themselves.

 

Log in or Sign up to hide this ad.

I am. The Proof:
You are here.

 

You are forgetting your youth.
Maynard never used the G. Dobie never used the G when he introduced him.
Maynard always signed Maynard, because the G was for his middle name Walter, because his mother didn't spell good.
Sooooo.....

 

Then would it not be an AI subjective opinion?

 

Excellent point. There's a lot variability that wouldn't be seen with just a sold price!

What about consciousness c;

 

I'm still waiting for a convincing proof that I'm conscious.

 

Dood, let’s see it beat a professional at tennis. It can never do it or anything for that matter that requires feel.

 

That seems like a bad example. Mechanical motion control and quantitative computer vision (get precise location/trajectory from images) have been superhuman for a long time.



How many sets would a human player last against a robot perceiving and moving at those speeds?

 

You can’t program human emotions in these things. You still don’t get it. You can make it vastly superior to us, but only intellectually. But when that 100 mph shot is coming at you 1 inch above the net, what do you think you were doing when you returned it 1 inch in bounds from the back line at 110 mph for the winner, do you think you were thinking, calculating, intellectualizing? You were emotionally reacting. That’s a different mechanism. It’s not intellect, it’s your emotions that did it. You can’t program the latter in AI, nor train it to have it. You can program to simulate it has emotions, cognitively, BUT THAT’S IT. Fake it. Sort of like what a sociopath does. Capisce, there, paisan?

 

I think its hard, if not impossible, to teach the AI to grade coins as good as people, if even people themselves cannot agree on coin grades.
I mean, everything is peach and perfect until there are very standard looking coins. Maybe it would be possible to let the AI to grade silver eagles, these are almost all in perfect or near perfect grade, and usually the decision is 70, 69 or 68.
As soon as there more complex decisions, the problems appear.
How should the toned coins treated? I have understood, that years ago the toned coins were treated as damaged coins (environmental damage), today they are mostly accepted. Maybe in 20 years these are treated as damaged coins again? I have understood that for example CAC is very harsh on these, most get artificially toned designations.
Or what should be the grade, if the coin is weakly struck?
Or what should be the grade, if the coin is struck with unevenly worn out dies (for example on side has been struck with fresh die, other not so much). Or coin has just uneven wear, has been in jewlery for example?
Or medieval coins. They might be bent or unevenly struck because of the manufacturing process.
Or what should be the grade, if there are chopmarks? Counterstrikes?
Or mint errors? Clips for example?

During the history, literally hundreds of thousands (if not even millions) of different coins have been made. The AI should be able to recognize all of them to understand, what technology has been used to make them, what part of the coin fields is scratches and which part is actual design elements and so on.
I wont say, that its impossible to teach the AI to deal with all of this, but how much time does it take to get more or less accurate model and what does it cost? How much overhead would it add to the current business models? In short - would it actually make sense? Would it pay off?

But I can see some potential in AI in some simpler steps. It can maybe help the TPGs (or actually graders) to avoid technical errors. For example the AI can verify, that the coin in the slab is actually dime, not quarter. That would save TPGs some money, less reslabing. Or maybe it can detect potential problem areas on coin, and the grader can then decide, if this is actually problem or not. Or maybe it can warn the grader, that some type of coins are known for fakes, and can point out some known tell signs.

But general AI, which can detect coin types, varieties, fakes, assign grades and write Shakespeare poems at the same time? I'm a bit skeptical about that part... Usually the tools that are meant for everything are not good at anything. But I will happy if I'm wrong

 

LOL. My paranoia indicator just ticked up a notch. Gonna have to keep my computer locked up at night with a good old fashion padlock.

 

Good one, Fido. Some light reading, now, just for you…

“Imagine—if you can—not having a conscience, none at all, no feelings of guilt or remorse no matter what you do, no limiting sense of concern for the well-being of strangers, friends, or even family members. Imagine no struggles with shame, not a single one in your whole life, no matter what kind of selfish, lazy, harmful, or immoral action you had taken. And pretend that the concept of responsibility is unknown to you, except as a burden others seem to accept without question, like gullible fools. Now add to this strange fantasy the ability to conceal from other people that your psychological makeup is radically different from theirs. Since everyone simply assumes that conscience is universal among human beings, hiding the fact that you are conscience-free is relatively effortless. You are not held back from any of your desires by guilt or shame, and you are never confronted by others for your cold-bloodiness. The ice water in your veins is so bizarre, so completely outside of their personal experience, that they seldom even guess at your condition.” -Martha Stout, Ph.D., The Sociopath Next Door.

 

Although at some point in the future such an accomplishment may be viewed as a victory in human terms; scoring a single point in a tennis match does not a victory make.

 

Speaking of the “future” AI will never be able to beat a human being in something like that that requires emotions because not having any emotions it will never be able to play as good as a human being. It can learn to beat us in hopscotch but not in anything that requires emotions to excel. And it is surprising to me not one of you seems to grasp that.

 

John Henry went to his grave with hammer in hand believing the same.

 

I’m talking about what AI can and can’t do. What are you talking about?

 

Just like you - I'm talking about what AI can and can't do.

 

We disagree on what tasks do and don't require emotion.

We probably disagree on whether machines can ever have emotions, too, but that's neither here nor there.

 

Being that AI has no emotion it can't be beat because it doesn't care whether it wins or loses.