Will Compu-grading eventually work?

oldcameoproofsguy · January 30, 2003 2:16PM

Will machines ever be able to grade better and far more consistently than people? Technical aspects and strike are easy but can something like eye appeal and luster really be programmed into a computer.

I believe it can. I think Compu-Grade failed simply because the programming and scanning technology was seriously lacking. IMO, machine grading will eventually far exceed the "art" of grading.

braddick · January 30, 2003 2:21PM

I honestly don't believe you can ever program a computer to think and feel like a human.
And, without being able to do so, will not invent a computer that will accurately grade.

It's not just a science- it's a feeling.
Computers don't have feelings.
(Of course, neither do some graders, but that's another Thread. . .)

hughesm1 · January 30, 2003 3:19PM

From a technical standpoint, YES
For "market grading," NO

Too many human variables...blast white v. monster toned etc.

michael · January 30, 2003 4:01PM

witrh all the factors it takes to gradce a coin all combined

NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NONO NO NO NO

ABSOLUTELY NOT took me 35 years to get where i am today to look at only a few coins in my specialized field that i love and have been doing this 24/7

sincerely michael

you need there things

a good glass
a good mind
and at least 20 years of looking at coins! maybe more!!

MastaHanky · January 30, 2003 4:05PM

If you're looking for strictly technical grading, a computer will trounce a human all day long. And its accuracy will be much better than a human's.

But like the others said, it won't be able to compensate for eye appeal.

Cam40 · January 30, 2003 6:51PM

A machine that can grade coins is only as good as the info put in the machine to come up with a grade.
What if the programmers were AGC graders?Oh brother.What a nightmare.

CharlieB · January 30, 2003 7:44PM

<< <i>If you're looking for strictly technical grading, a computer
will trounce a human all day long. And its accuracy will be much
better than a human's. But like the others said, it won't be able to compensate for eye appeal. >>

I think there will be a good attempt made as more technology is
applied to collecting. It's amazing what types of devices are
popping up like the picture of the HP scanner in the Wall Street
Journal yesterday that looks at a stack of old prints or
negatives for your PC. I am in the computer business and with the
progress I have seen in technology nothing would surprise me at this
point.

Click on my website and the link "Futuristic thoughts on Coins" which
seems to be tailor made in response to your thread. But do I think it
will satisfy collectors. Big question mark ??????

greghansen · January 30, 2003 8:29PM

I think computer grading might actually be usefull for 'commodity coins' and the sight unseen market. It will never have a chance for the special coins, however, other than to perhaps set a technical grade baseline.

CLASSICS · January 30, 2003 8:46PM

and who will program the computer?, will thier sofeware always stay they same? or will someone come along in a few years and change it? you can never remove the human eye or element from grading. there is, and will always be a diffrence of opinion on grading.

dwood · January 30, 2003 8:57PM

Computers will eventually be able to grade more accurately, and certainly more objectively than humans. The only question is "when?" Like voice recognition, adaptive reasoning technology is in its infancy, and years from fruition, but it is coming. Computers will eventually be able to study data, and reach their own conclusions based on past results. In terms of coin grading, I anticipate years of debate over "which is better" ultimately ending in an acceptance of a computerized standard. I also anticipate being dead and gone long before it happens

It isn't a question of "if," its merely a question of "when?"

nwcs · January 30, 2003 8:59PM

I think computer grading can only go so far. So much of a coin's grade currently is based on factors that cannot be quantified. Now, if we agreed to only grade a coin based on quantifiable factors then computer grading is easy. But the prices would never reflect the grade.

agfox · January 30, 2003 9:00PM

Yes, some year down the road. I believe it will be developed for something else first, then adapted to coins. It's all opinions, different graders have different opinions on coins, so what's different if a computer has one too? As long as the software is developed enough, it'll happen. Graders now swear they can grade a coin on average, under 15 seconds. There's a lot going on upstairs in 15 seconds. To a computer that's an eternity. Just give it time... Then we'll all have something else to b*tch about

Barry · January 30, 2003 9:37PM

I can see it now...
Everybody complaining about the computer virus that hit Accu-puter Grading. Made it overgrade everything...

Coqui · January 30, 2003 9:41PM

I really dont like compu-grade..... I always said there is nothing better than a good magnifier and a good light. I dont trust to much compu-grades... sometimes they are based on some base pictures and are not exact..... I tell from my point of view

ccrdragon · January 30, 2003 9:54PM

The biggest problem is that eye-appeal is totally subjective - case in point toned coins. I personally do not like toned coins and prefer to have mine either blast white or red (cents) but there is obviously a real market for toned coins today. How do you program a computer to say that this particular or that particular type of toning is beautiful and another isn't?

dwood · January 30, 2003 10:05PM

The same way you teach yourself that "this particular tone is pretty and that one isn't." When you think of it terms of your own personal decision, it seems simple...either its pretty or its not...you just know right away. But the fact is that subconsciously, you're wheeling through thousands of memories, prejudices, reference points to arrive at that decision. If a group of people can agree on the grade of a coin (and obviously we can, or else there would be no accepted grading sysytem), then a computer can be "taught" the same system. When it comes to eye appeal, again, it will still come down to "beauty is in the eye of the beholder" but the possibility of teaching a computer to identify "desireable" levels of toning, or shades, or patterns, or whatever, exists. And its only a matter of time and technology before it is implemented.

DO I look forward to that? No. I like there being some flexibility and opinion in grading. Do I dread it? No, because there will always be an element of "well, I just like this one better." I have several coins with grades lower than other specimens in my collection which I would never part with. They just look better to me. Technically, they're lesser quality. Aesthetically, they're superior. But that's a preference.

For example, I defy you to show me a pretty Buffalo nickel. There is no such beast. In my opinion, it is the ugliest coin ever minted anywhere, in any country, at any time in history. But that's a personal preference. A computer would have no such preference. A computer could see that it is an aesthetically pleasing coin (to some), and could/woould grade it accordingly. To me, there all culls

kranky · January 30, 2003 10:20PM

Won't work.

1. Services would have losses because there would be no more crackout/resubmissions. Computers would grade the same coin the same way every time.
2. No two coins are alike, so the computer would have to make assumptions instead of a straightforward scan/process/grade operation.
3. Market grading looks at the coin, figures what it would sell for, then assigns the grade which corresponds to that price. Can't do that with computers.
4. We assign MS coins to 11 different grades which is beyond the ability of people or computers to reliably determine. It would be like having a machine that can measure to within 1 millimeter, and then trying to use it to measure lengths of 0.25 millimeter.
5. No matter how you process the data, the initial input to a computer grading process would be a picture. That means people could play with coins in such a way to alter the way it would appear in a picture.
6. In order to see hairlines, the coin would have to be photographed from many angles. It would be faster to do it by hand.

OneYearTypes · January 30, 2003 11:51PM

Why not store pictures of every existing high grade-low population coin and have the database search and compare the coins against each other?

dwood · January 31, 2003 1:25AM

"OneYear" is onto it. And I suspect that is exactly how the process will begin. After that, it would then be a logical progression to have intuitive reasoning processes that would construct an an ideal example and then compare results against that. With each example grade, the ideal would be altered slightly, as a composite of all specimens graded.

As for photographing from several angles, the technology already exists to photograph a coin from an almost infinite number of angles in a fraction of a second (literally).

We won't see it in our lifetime folks, but you can bet that an accurate, intelligent, adaptive computerized grading process is on the horizon, and will one day be the norm.

airplanenut · January 31, 2003 1:30AM

Suppose it could see wear and contact marks- that still doesn't cover the problems associated with such things as weak strikes... only a person will know exactly what to look for for specific dates/types. Also, alterations are easier, such as minute retooling or slight cleanings/artificial toning.

Jeremy

nwcs · January 31, 2003 1:37AM

OK, let's look at it like Commander Data on Star Trek. He's what you're describing as the perfect coin grader. According to one episode, he basically said he can analyze the aesthetic qualities of an item against known norms but he cannot say whether it is appealing or not. Only what he thinks *should* be appealing for someone. Like his poetry. The same will be true of computer grading. It can give us some analysis of aesthetic qualities but will never be able to tell whether it is pleasing or not. That requires emotion.

MICHAELDIXON · January 31, 2003 1:53AM

This is a question David Hall could answer with some background information. PCGS was working on computer grading in the early 1990s.

keets · January 31, 2003 2:00AM

........it's reasonable to assume that a system can be devised with a program that would judge coins against a known, theoretical, perfect specimen and assign a grade accordingly. with that technical grade, a grader could then grade the coin asthetically for eye appeal.

that in itself is a powerful statement about todays grading companies. if the right person looks at our coins, we get a better grade based on his/her likes/dislikes from an asthetic point of view. with a properly developed computerized system, the technical grade would at least be consistent and beyond reproach. that would cut down on whining by about 50% when grades are posted here, leaving only complaints on eye appeal, which are of course very personal.

in the end, some type of human involvement will always be needed, at least until AI with human qualities is a reality. but by then we'll probably not be using coins, so we'll all get the discounted re-grade fee when we submit!!

al h.

dwood · January 31, 2003 2:03AM

True.
But what is pleasing to you may not be pleasing to me (this was the point of my reference to Buffalos above). Computerized grading would not eliminate emotional premiums anymoreso than what we have now. What we have now is an imperfect system where we encourage, promote, and foster ambiguity and personal interpretation. What a standardized system would give us is a base upon which that ambiguity could be further emphasized (we KNOW its an MS67, now...how much do you LIKE it?)

Lacking emotion is often a good thing. Computers have no desire to label an MS67 as an MS69. There's nothing in it for the computer

If its an MS67 that has beuatiful tone, then I'll pay a premium for it. But I don't want to call an MS67 an MS69 just because it has some pretty colors. In reality, its just a really pretty MS67. So start the bid at MS67 money and lets see who thinks it is prettiest.

nwcs · January 31, 2003 2:09AM

Really, it's the same thing as going back to technical grading. As I said, breaking the relationship between grade and price. The market didn't go that way, though. It's like computers and client-server setups. Every 10 years or so the industry goes from decentralized to centralized and back. Seems no one has figured it out and the next big thing shifts the industry in a direction till the next big thing shifts it back.

Nothing new under the sun.

Neil

keets · January 31, 2003 2:16AM

hey d

hence my middle paragraph. much of the problems today are simply a grader liking a coin or not. that's a real world function that takes place whenever i look at coins. the other night i paid $9 for a 1972 mint set at a club auction and heard the guys behind me mumbling that the set lists at $3-$4 greysheet, tops. i just turned around, smiled, and explained to them that grey sheet is average and the sheet hadn't looked at the set i had just bought, which had a 65FS P mint jefferson. i also explained to them that there were some sets i wouldn't pay $3 for. just like a computer that would average a coin, sheet prices are rather impersonal. human involvement is always neccessary at some point. i just hate it when that point is me returning a coin i bought sight unseen!!!

al h.

cladking · January 31, 2003 4:56AM

If there were grading standards a computer could grade better than a human now. Sure there
would be major problems with such a system because of it's inability to detect problems and at-
tributes that a human grader might spot at a glance, but for the majority of coins it would be more
accurate and consistent. Computers still have a hard time "seeing", but this technology is advancing
rapidly. Within a few years computers should be able to handle the bulk of the grading chores and
it may not be so very long until they can do it all. There will be electronic intelligence in the future,
but this won't be required to get consistent grading.

nwcs · January 31, 2003 5:06AM

But how well would a computer handle intuition cases like really well done artificial toning or where there is confusion between a low mint state weak strike and AU?

Bear · January 31, 2003 5:27AM

I suppose that the ultimate end of the road ,would be for robotic collectors,

buying coins from robotic dealers and graded by robotic machines.Thus we will have a perfect

closed circle, with all in total sync with the hobby. What a really peachy future to look foward to.

People are really such a pain in the a$$. Bear

clackamas · January 31, 2003 5:29AM

The question was "will it eventually work", and given that there was no time frame the answer is a certain, Yes. 50 years from now computers will be so much more advanced. You will probably be able to have a 200 DPI tiff image of every coin ever graded stored in the 100 TB of real memory the computer has. It could then search the current coin and come up with a precise grade. The system will also be able to learn and as it is used more become better. This will be childs play 50 years from now and you will be able to buy your own system for a grand.

cladking · January 31, 2003 6:30AM

<< <i>But how well would a computer handle intuition cases like really well done artificial toning or where there is confusion between a low mint state weak strike and AU? >>

A computer doesn't need to have an opinion on toning. Two people can't agree on
toning so why would a computer. It could simply state the wavelenght of the reflected
light and it would be the collectors opinion that mattered.

It would need to have parameters for the amount of luster which can be disrupted be-
fore being downgraded. Remember though that with true standards there would be
many more grades, so downgrading in a single category would not be so dramatic.

cjmbuffalo · January 31, 2003 7:09AM

I agree that its just a matter of time before the computer can do the technical task. The bigger question is why not change the grading system at the same time. Design a system that works with a computer. Who says that the grading parameters cannot change?

cladking · January 31, 2003 7:45AM

<< <i>I agree that its just a matter of time before the computer can do the technical task. The bigger question is why not change the grading system at the same time. Design a system that works with a computer. Who says that the grading parameters cannot change? >>

Why not start in that direction now with something like a strike designation?
It won't be as big a shock later.

nwcs · January 31, 2003 11:25AM

<< <i>Why not start in that direction now with something like a strike designation?
It won't be as big a shock later. >>

Like I said, this goes back to the days when grading was technical and not market driven. But even then, there were considerations for market appeal. The marketplace apparently ruled that technical grading was insufficient. Otherwise we'd be technical graders. But who knows? Maybe the "next big thing" will swing the pendulum back till the "next big thing after" swings it around again.

There is nothing new under the sun. What has been will be again.

Neil

oldcameoproofsguy · January 31, 2003 1:01PM

Many have suggested that a viable computerized system is 50 years down the line or not in this lifetime.

If one looks at processor speed versus the time needed to develop processor technology, in 20 years processors will manipulate information as fast as the human brain. This is assuming that processor speed continues to grow at the present rate.

I think that an eloquent grading machine will be here in the next couple of decades.

I feel it will also change the way that grading companies do business but not kill a grading companie's business. The slabbing process will become cheaper so people will send in less valuable coins. It will also better record rarities in the market place; all the real rarities could in essence be ranked.

Finally, I feel that such a system could learn through previous grading what constitutes eye appeal, luster, and strike. It could also easily be programmed to take into account market value into grading. It could probably eventually be better than the human eye at determining AT (sulfur and other compounds emit specific resonance frequencies that a human eye cannot detect). It would also be able to determine forgeries. Most important, it would be completely unbiased.

kranky · January 31, 2003 8:34PM

I doubt that slabbing would get cheaper. The only part of the entire process (open package, verify coins on invoice, enter coins into computer, store in box, grade, package in slab, quality check, pack and ship) that the computer would do is the grading, and that part takes under a minute. Add in the need to cover the cost of developing a computer grading system, and I wouldn't be surprised if costs actually go up. If, for the sake of argument, the computer is better, they could charge more because it would be a better service.

I'm pretty familiar with technology (started working with computers way back on a DEC PDP/11 minicomputer) and even though imaging techniques will improve, computers will get faster and storage cheaper, the technique of grading coins is just not a good match for what computers are good at. Some of you guys are familiar with programming - imagine how much code it would take just to determine if a Jefferson is full steps or not.

If computer grading requires a large database of images to be successful, imagine the cost to develop it. How many very detailed images would be needed for each series of coins, for each commemorative design, etc. just to get started? I'm guessing a lot. Now we have to program in the "adjustments" - if a valuable coin is 10x the value in 65 than it is in 64, the system would have to make sure the coin is a solid 65 and not a just-made-it in order to get the grade because that's how it works with human graders.

Maybe it could be done with moderns (little chance of counterfeits and altered coins, most coins that would be sent in are MS65-70, and you could get started with a lot fewer images required), but I just don't see it as possible on a wide scale.

MacCoin · January 31, 2003 8:44PM

not in my life time but I don't know what will happen 50 years from now.

matteproof · January 31, 2003 9:14PM

I prefer to keep our human graders EMPLOYED and I would never opt for computer grading as a "stand alone" system. Human beings are good enough (some are great) at the job of coin grading. Such a computerized system would contribute to the inevitable, deplorable and UN-AMERICAN firing and lay-off schemes (as is so common today) that would surely follow. No coin grading system is worth contributing to that kind of garbage. In many ways, computers are a mixed blessing. I perceive that they take away as much (or more) than they actually give.

matteproof

cladking · January 31, 2003 9:51PM

Grading will always be a valuable skill to have. Some one has to select coins to send in,
and the computers will need a human to operate and baby sit them until machine intell-
igence is invented. There are very few professional graders anyway, if you want to worry
then worry about programmers.

58Flyer · January 31, 2003 10:14PM

ABSOLOUTLY NOT!

First-We can see the grade, mint marks and all showings much better then some dumb old robot.

Second-We were the ones who built computers so they would never be as good as the makers.

Third-Computers don't even have feelings and can only rely on its mechanics.

Fourth-Computers might be able to find dates and colors (if it were toned) but it would never be able to find hits and rubs, thats where Humans come in.

Will Compu-grading eventually work?

Comments

Leave a Comment