How long before coins are no longer graded by people but by computers?

It sounds like you're saying that scanning a coin damages it. Of course, that's not true. If you put your coin on a flatbed scanner and scan it, you'll get an image that doesn't show the coin's luster.

Machine grading would use other scanning techniques, probably involving a number of different scans for each coin.

 

Log in or Sign up to hide this ad.

I think that tells it all. Oh - people still won't be able to grade like desertgem said.

 

Hahahahaha! This just made my morning

 

And those patents are about 23 years old now and will expire in just a few years.

 

I, personally, do not think so. I'm thinking 100 Thousand Years since computers can calculate probabilities but you just cannot program "like"!

AND, unfortunately, "like" dictates price and price = grade.

Folks have tried for a long, long time to "define" a specific grade in absolutes. Dings per square mm, number of hits, depth of hits, acceptable or unacceptable abrasions, luster, lack of luster, full or weak strike and on and on and on. These values simply cannot be equated nor absolutely defined which is what a computer would require. All the different combinations of the above attributes adds even higher levels of incalculable complexity.

As such, it all boils down to what folks like vs what folks find acceptable vs what folks find unacceptable. Those three items equal price, according to current market conditions, which have their own levels of complexities.

In coin grading, to put it simply, there are no absolutes.

 

Computers do not think! They are NOTHING more than a simple adding machine.

All the code in the world breaks down to a simple 0 or 1. On or Off. Basic simple comparisons with no levels of complexity other than instruction quantity (i.e. how many times per second can 0 be compared to 1?)

Some folks may wish to believe otherwise but it doesn't make it so.

Yes, technology has advanced but not to the point of enabling a computer to actually think.

If they did, they certainly would not put up with our emotional inadequacies for very long since most emotional response is simply not logical.

 

I am going to have to try that when I get home.

 

The day that happens is the day coin forums cease to exist. What else do we have to bellyache about other than bashing the grading companies?

Yes, I see it coming, we have retina identification, we can run computers thought processes. Just a matter of time

 

I think it will happen within the next ten years, at least partially. It promises be the solution to lot of the inconsistency we experience today within and between TPGs. We'll also know exactly why a coin is a MS62 and not a MS64, in precise terms, and eliminate costly resubmissions in hopes of getting a better grade. It should help with counterfeit detection if photo analysis is combined with weight analysis, magnetics, and spectrum analysis. It might be that the technology already exists -- or am I watching too many CSI shows on TV?

 

Once someone comes up with a program that compares all of the coins perameters to the perfect example then a computer could grade a coin. Of course the scanner and scanner method would have to be the same for every coin grades by GCGS (Guaranteed Computer Grading Service) . I can see it now MS68.764

I don't think you would see and MS70 ever again, but rarely would 2 coins grade the same exact grade if they went out to 3 decimal places.

Lack

 

That was what they said about grading our children's intelligence and educational progress. If you want computers to grade coins (like they already have been put in charge of grading our children), all we have to do is value the things that CAN be graded well by computers (spelling, names and dates, number and size of bag marks) more than those intangible "qualities" (the ability to make and achieve goals, the desire to learn more, "eye appeal").

Sure they may start with the lower grade coins (or students), but eventually we will adapt to this new "better and more consistent" way to judge things in our stead. It will eventually become so normal that we will crave it; we'll call it "getting back to the basics of coin grading."

 

How will the machine be able to differentiate between wear and a weak strike? For instance, 1921 Peace Dollars have a high relief and are often not fully struck up which makes uncirculated examples appear to have circulation wear to the untrained eye.

 

Grading coins involves more than just ones and zeros. Asthetics involves knowlege of varying degrees of quality (or lack thereof) and emotion or opinion. You can create a program that will look for brightness, scratches and definition all in 3D. Could you program a computer to differentiate between a previosly unknown die variation and PMD? Better yet, can you get a computer to understand the emotional aspect of the definition of beauty? That's the big one!

 

I was using the word "think" to keep things simple and imply several different modes of data processing. Perhaps it would have been better if I had put it in quotes, but I was pretty tired and didnt think of it.

I believe you're saying what I meant about the lower grade coins, that they might be quantifiable by computers at this time, but only for simple tasks like sorting them to be looked at later by a person. Or rather, sorting out grades with a few general categories. To judge a coin well, especially a higher grade, would seem to require a sense of aesthetics, and it'll be a while before computers have that, IMO.

------------------

But really, theres no way to tell. It might be possible for a really clever programmer to figure out a way for a coin to be judged with a great deal of accuracy with near current technology, or it could require the development of limited artificial intelligences which I believe are still pretty far off--hardware like the hypothetical neural net thats supposed to emulate the way a human brain thinks.

 

Not only is grading by machine inevitable it is possible today. All of the equipment necessary to do what is done by the eye of a human grader exists already and is being widely used in other industries. All that remains is for the technology to be adapted to coin grading.

Exact surface topography can be scanned by a laser (which I believe secure plus does now). Luster can be measured by surface reflectivity. A mass spectrometer can determine composition (which I think the PCGS sniffer is doing now). Spectrophotometry can even measure the toning on a coin.

The process will begin with a coin coming to a grader and within 5 or 10 minutes it will have moved through several machines and collection of quantitative data about that coin will exist that no human ever has come close to having. Now the machine will assign a grade and it will not have to think at all to do it. All it has to do is mindlessly mimic the decisions previously made by a large group of human graders. Imagine 100 graders grading 100,000 morgans and all of those coins go through the same comprehensive data collection. Now that the data set exists it is easy to take the human grade and see what quantifiable data is common to all graders for a given grade. From there it is a small step to creating a program that can predict how a human grader would evaluate a coin. You don't need to evaluate if it is wear or a weak strike, all that is needed is a large enough sample human graders who have done that in the past and the machine can compare and predict.

If that process sounds unlikely it isn't. It is being done in astronomy now. See a web site called galaxy zoo which is using the general public to classify pictures of galaxies and is then using that data to train computers to do the same.

At first machines will recommend a grade and a human will confirm it (the PCGS sniffer partially does this now). Then the machines will issue a grade and a human grader will be there to reject it. Finally a machine will be the only grader of a coin. Human graders will still be around for things like super rare coins for which a large enough data set does not yet exist and new issues and things like that but the bulk of submissions will be done completely by machine.

Machine grading is not only coming but the transition process has already begun with the PCGS sniffer. A coin rejected by the sniffer comes back as what? Altered surfaces? That is a grade in itself and is being done by machine. It is only a matter of time until all facets of grading can be reproduced by machine. Someday you will see decimal grades and it wouldn't surprise me if grading went to several decimals.

 

Good post, but I just disagree. Yes, machines very well are capable of this. However, the problem is computers cannot replicate how humans perceive a coin. Even professional graders differ on this. Now, to grade ASE's, modern commemoratives, and the like, of course this could be done fairly easily nowadays. However, I challenge the assumption a computer could take a batch of capped bust halves, sort out those classified as detail grades, and evaluate these coins with and without toning, all identically to how a human would. I agree a computer could be trained to do so, but the downfall of this idea is that humans are the audience for the final grade, not other computers. There is more to grading than simple mathematical computations of luster, surface topagraphy, and mass spectography readings.

If its not more than these things, and in the future our hobby is filled with MS62.875428 coins with mathematical computations of exact pricing based upon the assigned grade, then I am out. If I am simply dealing with a bunch of numbers for numbers sake, I will stick to the stock market and make money, as my "hobby" will be no more than a busman holiday for me. I collect coins for art, beauty, and history, not for a number to plug into a spreadsheet.

Chris

 

Wear is determined by low relief combined with a lack of luster. A weak strike has the low relief but also retains some luster.

This would be very difficult to program.

Something else that would be difficult to program is the difference between die polish lines and scratches since technically, bith are scratches but one is in relief while the other is in the surface of the coin.