The R&D and Marketing involved would be orders of magnitude higher than that of a normal human eye grading service. I think most would agree to that.
A limited and stagnant market
How many active submitters are there? Is that number likely to grow in five years, ten years, twenty years?
Expected Return
will it be profitable? where is the break even point?
Longevity
If this works then the majority of coins worth grading will be graded with the non changeable grade within a few years. Customer base will shrink as material with non changeable grades decrease...
This is a good idea but it looks to be self defeating to me.
Remember this is a fantasy! You have posted good reasons WHY THERE CANNOT BE A SET (never changing) STANDARD and why computer grading will not be done. What can be done is computer- assisted grading but the same problems will be present.
More than any start up costs has already been burned trying to grade w/computers. From what I understood, a "grader" was assigning the "computer grades." LOL. Computer assisted grading could start tomorrow at PCGS without the "group" of experts. They could manage with the folks they have. So, strart up cost: Very minimal.
No R&D needed, computer "fingerprinting" is already around.
and 4. IMO, all TPGS will be hurting for money in the future. I predict that one popular service will be gone in another year or two. They are hurting now and you can only lower prices so much.
Remember, any coin worth grading...
Eventually, the coin will get into the correct holder and that grade will not go higher UNTIL the gold posts are moved again and the whole "show" begins again until its "new, modern grade" is achieved. We don't need computers to do that.
@chesterb said:
As I've stated before which might have been lost in the discussion, a computer most likely (probably 95% of the time,) diagnosed your loved one's mammogram before the radiologist even sees the images. It highlights areas that it believes are abnormal and provides an initial diagnosis. I don't think it would be difficult to duplicate a similar thing for coin grading with a finalizer acting like the radiologist and confirming the grade/diagnosis
I've seen things that computers are doing now and seen presentations by futurists. Any routine work done by a human will be taken over by computers in the future. Its already happened in some industries. Who would have thought of computers driving cars but that will happen!
Nothing was lost to me in your post. It was well stated. Unfortunately (?), coins don't have breasts. Therefore, until a computer is finally given the ability to cry, enjoy the beauty of a sunset, learn the value of a dollar, and be programmed with the current dictates of excellent taste and eye appeal, that robot won't work.
@chesterb said:
As I've stated before which might have been lost in the discussion, a computer most likely (probably 95% of the time,) diagnosed your loved one's mammogram before the radiologist even sees the images. It highlights areas that it believes are abnormal and provides an initial diagnosis. I don't think it would be difficult to duplicate a similar thing for coin grading with a finalizer acting like the radiologist and confirming the grade/diagnosis
I've seen things that computers are doing now and seen presentations by futurists. Any routine work done by a human will be taken over by computers in the future. Its already happened in some industries. Who would have thought of computers driving cars but that will happen!
Nothing was lost to me in your post. It was well stated. Unfortunately (?), coins don't have breasts. Therefore, until a computer is finally given the ability to cry, enjoy the beauty of a sunset, learn the value of a dollar, and be programmed with the current dictates of excellent taste and eye appeal, that robot won't work.
With substantial development cost, a system could be created which would output the same grade every time,
if the coin itself did not change, and if the weights applied to the different observable attributes did not change.
The weights could be fitted to coins already graded.
Would it be good enough if it agreed with the human grade 75% of the time? 90%?
It gets tricky, because we know the human grades have a nonzero variance.
In theory, you could develop templates for each obverse and reverse type which define the prime focal fields.
Rotate the coin until the template lines up.
Then count contact marks and size of marks in those areas on untoned coins.
It would require more than just a TrueView - you would want light from many angles - maybe 9 angles?
All of these tasks have some complexity.
For coins with toning, then you have the problem of distinguishing between a contact mark and a place where the toning changes, so more complexity.
How to define attractive toning? Maybe possible, but seems complex.
The machine could be good for measuring % Red on copper.
But a photo of the coin will communicate this info to a buyer very well, if they care about small differences in % Red.
Looks like a lot of cost to me, not really balanced with laying off a couple of graders.
https://youtu.be/mr9kK0_7x08?t=145
(at 2:25 especially)
As Elon Musk learned on the Tesla assembly line, in theory you can automate everything,
but humans are really better at some things (like manipulating small parts and doing complex things like grabbing
a dangling wire and connecting it). And humans are not that expensive to pay.
At some point in time in the future the current business model of Third Party Grading will become unsustainable. At that point an innovator will surface who will propose something to fix what ails the hobby. This proposed fix will be evaluated, embraced, lauded and skillfully implemented with great success. This new fix will become the accepted norm, until it too falls out of favor, only to be replaced again.
It would be interesting to see if the new fix consisted of promoting the idea that the hobby will be best served if all coins were raw (I assume someone could come up with valid reasons why and persuade all segments of the hobby to follow along).
If this new fix takes hold, crackouts of coins to liberate them from their plastic coffins would become the norm. Collectors and others in the hobby could again (or maybe for the first time!!!!) experience the joy of holding their favorite coin in their hands (by the edge of course, over a velvet tray).
Then when most coins have been so liberated, there would be a vast ocean of raw coins to tempt the next innovator. That ocean of raw coins eventaully will be seen to be in dire need of being professionally "graded" and protected by a new and improved container.
None of us has likely seen state of the art AI (I know some of you probably have very high security clearances, so I might be wrong on that). The technology to do this does already exist. Take your 10 experts, an ultra-high res scanner, and a super AI. Have the AI grade over the experts shoulder for 6 months and it will grade exactly as they do...tarnish, marks, biases, and all. And the high res scanner will know every time the same coin is submitted, or will know if it's altered, and it doesn't matter anyway as it, unlike humans, will grade the same coin the same way every time.
Also, as to EAC being complicated, the basis is 3 quality levels Choice, Average and Scudzy...kind of sounds like coins being sorted into A, B, and C levels within a grade. Yep...complicated.
@neildrobertson said:
The obstacles for something like this are not technical. Many people have the ability to create a fully automated grading system. The greater challenge is to get the world to adopt this new system.
Have any names for the "MANY PEOPLE you know of?"
BTW, IMHO, if one of your "many people" come up with this mythical "magic" machine (MMM) that actually works good enough to be accepted, a major TPGS would probably want to purchase the rights to control/use it.
I mean many people have the ability in the sense that many people have developed optical systems that look for fairly complex patterns or shapes in a lab or industrial setting., with high repeatability. There is instrumentation that can measure things like surface roughness for PL designations. There are many people that are able to develop interfaces between the instrumentation and the grading software. That are no conceptual or technical obstacles to automated grading that haven't been overcome in other areas.
I think a computer grading system might be tough to program for some things like toning and different variations in eye appeal. I am imagining a revision in grading standards or a new system could come along with the automation. I think such a thing has the potential to be a step forward for the hobby. We tend to be a pretty conservative group, so I don't imagine those sorts of changes to get too much support or fanfare.
It's not futurism. It's just a matter of someone (or some group) wanting to do it and whether the economics are there.
@chesterb said:
As I've stated before which might have been lost in the discussion, a computer most likely (probably 95% of the time,) diagnosed your loved one's mammogram before the radiologist even sees the images. It highlights areas that it believes are abnormal and provides an initial diagnosis. I don't think it would be difficult to duplicate a similar thing for coin grading with a finalizer acting like the radiologist and confirming the grade/diagnosis
I've seen things that computers are doing now and seen presentations by futurists. Any routine work done by a human will be taken over by computers in the future. Its already happened in some industries. Who would have thought of computers driving cars but that will happen!
Nothing was lost to me in your post. It was well stated. Unfortunately (?), coins don't have breasts. Therefore, until a computer is finally given the ability to cry, enjoy the beauty of a sunset, learn the value of a dollar, and be programmed with the current dictates of excellent taste and eye appeal, that robot won't work.
Most classic U.S. coins do have busts.
Not any I'm remotely interested in except to look for friction!
@shorecoll said:
None of us has likely seen state of the art AI (I know some of you probably have very high security clearances, so I might be wrong on that). The technology to do this does already exist. Take your 10 experts, an ultra-high res scanner, and a super AI. Have the AI grade over the experts shoulder for 6 months and it will grade exactly as they do...tarnish, marks, biases, and all. And the high res scanner will know every time the same coin is submitted, or will know if it's altered, and it doesn't matter anyway as it, unlike humans, will grade the same coin the same way every time.
As I said, I could probably do most of this. The AI part is not complicated. It's training a classifier with supervised learning. The hard part is collecting all the data that a grader sees so that the system can be trained to produce a good result (data engineering), and then creating something that can reproduce acquiring all that data in an automated fashion such that it gets data that's sufficiently the same to produce the same result each time the coin is presented to the system (contraption engineering). Having "the AI grade over the expert's shoulder" needs to be put in more concrete terms, and that's the big challenge here.
@shorecoll said:
Also, as to EAC being complicated, the basis is 3 quality levels Choice, Average and Scudzy...kind of sounds like coins being sorted into A, B, and C levels within a grade. Yep...complicated.
My EAC Grading Guide is at home so I cannot refute your post with actual quoted from that book!
Perhaps, yo can help me. I have a PCGS Photograde XF 1794 large cent.
Obverse: Rim bruise 2:00. Small, deep rim nicks 3, 8, and 11:00. Small dark green corrosion lump on the "L." Several /4 inch scratches on the bust Eleven scattered, bag mark-looking hits on bust and in fields.
Reverse; same sort of stuff including a little weakness at the bottom.
I contend that six of the best copper dealers/graders in the country would not "hit" the same EAC grade (price for sale). That's why it is complicated. I'll also bet that any EAC dealer, TPGS grader, knowledgeable collector would get the coins "design condition" correct most of the time BEFORE the copper guys "Net" it down.
@shorecoll said:
None of us has likely seen state of the art AI (I know some of you probably have very high security clearances, so I might be wrong on that). The technology to do this does already exist. Take your 10 experts, an ultra-high res scanner, and a super AI. Have the AI grade over the experts shoulder for 6 months and it will grade exactly as they do...tarnish, marks, biases, and all. And the high res scanner will know every time the same coin is submitted, or will know if it's altered, and it doesn't matter anyway as it, unlike humans, will grade the same coin the same way every time.
IMO, it ain't going to happen the way you posted! What I can believe is that one day in the far future AI will help human graders be more precise with their opinions. Thus cutting down on disagreements in house and in the market.
I wasn't saying EAC isn't complicated...I was saying everything is complicated...11 MS grades x 3 letters is complicated, without arguing with Bill Noyes how much a 1/4" ding drops a grade vs. an 1/8" ding.
@shorecoll said:
I wasn't saying EAC isn't complicated...I was saying everything is complicated...11 MS grades x 3 letters is complicated, without arguing with Bill Noyes how much a 1/4" ding drops a grade vs. an 1/8" ding.
I guess some of us have a different understanding of "complicated." For me, those 11 MS grades are the easiest to teach!
In my experience, examining a coin (with no time limit and using the proper tools and lighting) and seeing EVERYTHING there is to see on it is EXTREMELY EASY and not complicated at all! It can be successfully taught to any one with an eye-for-detail. That is the BASIS for all coin grading - the first step - properly examining a coin. It provides a coin's "true" technical condition of preservation from the time it was struck. I think many of us can do that if asked. As I posted above: "I'll also bet that any EAC dealer, TPGS grader, knowledgeable collector would get the coins "design condition" correct most of the time..."
Everything after that OBJECTIVE, PRECISE, exam introduces each of the subjective differences that cause virtually all of the problems/disagreements regarding a coin's "acceptable, 80% correct, commercial market grade." These SUBJECTIVE, IMPRECISE factors include:
1.Value
2. Eye appeal
3. Net Grading
4. Ownership
5. Skill/knowledge of the grader of the series, country, type, etc
"My friends who see my collection sometimes ask what something costs. I tell them and they are in awe at my stupidity." (Baccaruda, 12/03).I find it hard to believe that he (Trump) rushed to some hotel to meet girls of loose morals, although ours are undoubtedly the best in the world. (Putin 1/17) Gone but not forgotten. IGWT, Speedy, Bear, BigE, HokieFore, John Burns, Russ, TahoeDale, Dahlonega, Astrorat, Stewart Blay, Oldhoopster, Broadstruck, Ricko, Big Moose, Cardinal.
To pivot the subject a little bit, would a debate or panel discussion about the future of AI in coin grading be a presentation people would go to at a show, or would it just draw an audience of about 6, like many presentations do? For a debate, I would envision qualified people able to take both sides of the issue drawing their side out of a hat, not so much to convince people of what will happen and "win" the debate, but to lay out all the things to think about and provide 40 minutes of entertainment.
An assigned quantity representative of the detracting significance of each mark is then calculated by adjusting the measured surface area of the mark by a factor representative of the relative grading importance of the area on the coin where the mark is located.
Relative grading importance. Kind of a subjective thing is it not?
Whoever is careless with the truth in small matters cannot be trusted with important matters.
@mr1874 said: An assigned quantity representative of the detracting significance of each mark is then calculated by adjusting the measured surface area of the mark by a factor representative of the relative grading importance of the area on the coin where the mark is located.
Relative grading importance. Kind of a subjective thing is it not?
Not if it is pre-defined for the series. He's just saying that a hit on the cheek is more important than the same hit near the rim, or something like that.
I like this: "An assigned quantity representative of the detracting significance of each mark is then calculated by adjusting the measured surface area of the mark [and its depth] by a factor representative of the relative grading importance of the area on the coin where the mark is located."
Aha - was wondering if you were speaking as a professional grader or collector.
I know conflicts occasionally arise between the TPG's and owners of well known coins/collections and was hoping you'd have some inside tid-bits for us......
"My friends who see my collection sometimes ask what something costs. I tell them and they are in awe at my stupidity." (Baccaruda, 12/03).I find it hard to believe that he (Trump) rushed to some hotel to meet girls of loose morals, although ours are undoubtedly the best in the world. (Putin 1/17) Gone but not forgotten. IGWT, Speedy, Bear, BigE, HokieFore, John Burns, Russ, TahoeDale, Dahlonega, Astrorat, Stewart Blay, Oldhoopster, Broadstruck, Ricko, Big Moose, Cardinal.
The insights I gain here is helping with a new project on digital coin surface evaluation.
Please keep the descriptors and definitions coming!
HJP
Everything after that OBJECTIVE, PRECISE, exam introduces each of the subjective differences that cause virtually all of the problems/disagreements regarding a coin's "acceptable, 80% correct, commercial market grade." These SUBJECTIVE, IMPRECISE factors include:
1.Value
2. Eye appeal
3. Net Grading
4. Ownership
5. Skill/knowledge of the grader of the series, country, type, etc
@Lakesammman said:
Aha - was wondering if you were speaking as a professional grader or collector.
I know conflicts occasionally arise between the TPG's and owners of well known coins/collections and was hoping you'd have some inside tid-bits for us......
I cannot speak for professional graders. I've been allowed to use a stereo microscope to examine any coin I was grading at every TPGS I've worked at! I considered myself a "firewall" to detect counterfeits, repairs, and surface alterations. Ninty-nine percent of the time, I remove a coin from the box, get a naked-eye, gut reaction to its grade and authenticity with incandescent light, pop it under the scope using florescent light, and go from there. By then, most coins are graded and authenticated. Then I use a 7X hand lens under incandescent light to see what other graders/dealers/collectors might think (commercial grade). Then I grade the coin and leave notes. I've learned how some of the top graders see coins and have trained myself to closely match company standards. I saw some advice in one grading room: "Monkey see, monkey do, monkey get rich."
I don't deal with owners of well-known coins/collections. The majority of conflicts I've seen occurred long ago when big submitters complained about the grades on an entire order and threatened to take their business to another TPGS. I believe over the decades this has resulted in much of the gradflation that has taken place!
I do get to handle most of the customer relations that arise from grading or authentication questions. I get to play the devil's advocate and argue on behalf of our customers when they have a legitimate question. That's almost as much fun as looking at coins all day.
Yes, I think it's one of those many early Computer Age dreams that still hasn't been realized.
... and never will be.
Just like the MLB shouldn't get rid of umpires and use software/computer to call balls/strikes. It's part of the charm of the baseball and it helps keep fans and conspiracy theorists riled up and watching. Not much different that PCGS graders rendering opinions on coins. I'm sure each grader has their pet peeves and likes/dislikes that affect their opinion. That keeps submitters scared and excited to get their grades back. I'd think computer grading would reduce the number of coins sent in for grading which is not something PCGS would want to do.
@ricko said:
That patent was a glimpse of what the future must eventually become. It is possible, and, IMO, inevitable that the system will become reality, Objectivity must replace subjectivity if grading is to be respected. Just witness the disagreements we constantly see here....Witness the 'upgrades' we see constantly....Standards must become reality. Cheers, RickO
I think you are correct. But....how can a computer subjectively evaluate TARNISH? I suspect that you don't care. I know I don't!
I don't think that toning should affect the grade of a coin. It can affect the price of a coin, but not the grade IMO. People will say toning should affect the grade due to "increased/decreased" eye appeal, but that is subjective and dependent on the collector.
@neildrobertson said:
The obstacles for something like this are not technical. Many people have the ability to create a fully automated grading system. The greater challenge is to get the world to adopt this new system.
I don't think that a TPG would have to or would want to tell anyone who/what graded something. I'm sure graders wouldn't mind if a TPG start using AI to grade monster boxes of ASE's...
Interesting one. I think that some folks are underestimating the complexity, but it of course can be done. Is it worth doing? Not so sure. I've worked with computer vision, and it is definitely not trivial to detect marks on the huge variety of surfaces that coins present. The amount of programmer hours (which, like grader hours, are not cheap) required for this would be substantial. And the system would still need to be supervised for quite a while. Fortunes can be made and lost on grading points.
I think starting with moderns like (untoned) ASEs would definitely ease some of the difficulty. For classic coins, it could be super challenging, especially early era US where there is significant die to die variation. Computers don't see like we do, they are precise. If anything is out of place, a die break, etc, it can throw things off.
As a personal project, I once did some vision recognition on coin images. Even consistently getting orientation of the coin was challenging! And I was working in a single series.
Comments
Remember this is a fantasy! You have posted good reasons WHY THERE CANNOT BE A SET (never changing) STANDARD and why computer grading will not be done. What can be done is computer- assisted grading but the same problems will be present.
More than any start up costs has already been burned trying to grade w/computers. From what I understood, a "grader" was assigning the "computer grades." LOL. Computer assisted grading could start tomorrow at PCGS without the "group" of experts. They could manage with the folks they have. So, strart up cost: Very minimal.
No R&D needed, computer "fingerprinting" is already around.
and 4. IMO, all TPGS will be hurting for money in the future. I predict that one popular service will be gone in another year or two. They are hurting now and you can only lower prices so much.
Remember, any coin worth grading...
Eventually, the coin will get into the correct holder and that grade will not go higher UNTIL the gold posts are moved again and the whole "show" begins again until its "new, modern grade" is achieved. We don't need computers to do that.
Nothing was lost to me in your post. It was well stated. Unfortunately (?), coins don't have breasts. Therefore, until a computer is finally given the ability to cry, enjoy the beauty of a sunset, learn the value of a dollar, and be programmed with the current dictates of excellent taste and eye appeal, that robot won't work.
Most classic U.S. coins do have busts.
I bet it happens in the next 20 years.
With substantial development cost, a system could be created which would output the same grade every time,
if the coin itself did not change, and if the weights applied to the different observable attributes did not change.
The weights could be fitted to coins already graded.
Would it be good enough if it agreed with the human grade 75% of the time? 90%?
It gets tricky, because we know the human grades have a nonzero variance.
In theory, you could develop templates for each obverse and reverse type which define the prime focal fields.
Rotate the coin until the template lines up.
Then count contact marks and size of marks in those areas on untoned coins.
It would require more than just a TrueView - you would want light from many angles - maybe 9 angles?
All of these tasks have some complexity.
For coins with toning, then you have the problem of distinguishing between a contact mark and a place where the toning changes, so more complexity.
How to define attractive toning? Maybe possible, but seems complex.
The machine could be good for measuring % Red on copper.
But a photo of the coin will communicate this info to a buyer very well, if they care about small differences in % Red.
Looks like a lot of cost to me, not really balanced with laying off a couple of graders.
(at 2:25 especially)
As Elon Musk learned on the Tesla assembly line, in theory you can automate everything,
but humans are really better at some things (like manipulating small parts and doing complex things like grabbing
a dangling wire and connecting it). And humans are not that expensive to pay.
At some point in time in the future the current business model of Third Party Grading will become unsustainable. At that point an innovator will surface who will propose something to fix what ails the hobby. This proposed fix will be evaluated, embraced, lauded and skillfully implemented with great success. This new fix will become the accepted norm, until it too falls out of favor, only to be replaced again.
It would be interesting to see if the new fix consisted of promoting the idea that the hobby will be best served if all coins were raw (I assume someone could come up with valid reasons why and persuade all segments of the hobby to follow along).
If this new fix takes hold, crackouts of coins to liberate them from their plastic coffins would become the norm. Collectors and others in the hobby could again (or maybe for the first time!!!!) experience the joy of holding their favorite coin in their hands (by the edge of course, over a velvet tray).
Then when most coins have been so liberated, there would be a vast ocean of raw coins to tempt the next innovator. That ocean of raw coins eventaully will be seen to be in dire need of being professionally "graded" and protected by a new and improved container.
What a great hobby we have.
None of us has likely seen state of the art AI (I know some of you probably have very high security clearances, so I might be wrong on that). The technology to do this does already exist. Take your 10 experts, an ultra-high res scanner, and a super AI. Have the AI grade over the experts shoulder for 6 months and it will grade exactly as they do...tarnish, marks, biases, and all. And the high res scanner will know every time the same coin is submitted, or will know if it's altered, and it doesn't matter anyway as it, unlike humans, will grade the same coin the same way every time.
Also, as to EAC being complicated, the basis is 3 quality levels Choice, Average and Scudzy...kind of sounds like coins being sorted into A, B, and C levels within a grade. Yep...complicated.
I mean many people have the ability in the sense that many people have developed optical systems that look for fairly complex patterns or shapes in a lab or industrial setting., with high repeatability. There is instrumentation that can measure things like surface roughness for PL designations. There are many people that are able to develop interfaces between the instrumentation and the grading software. That are no conceptual or technical obstacles to automated grading that haven't been overcome in other areas.
I think a computer grading system might be tough to program for some things like toning and different variations in eye appeal. I am imagining a revision in grading standards or a new system could come along with the automation. I think such a thing has the potential to be a step forward for the hobby. We tend to be a pretty conservative group, so I don't imagine those sorts of changes to get too much support or fanfare.
It's not futurism. It's just a matter of someone (or some group) wanting to do it and whether the economics are there.
IG: DeCourcyCoinsEbay: neilrobertson
"Numismatic categorizations, if left unconstrained, will increase spontaneously over time." -me
Oh ...... the human factors ....... what would we do without it?
Not any I'm remotely interested in except to look for friction!
As I said, I could probably do most of this. The AI part is not complicated. It's training a classifier with supervised learning. The hard part is collecting all the data that a grader sees so that the system can be trained to produce a good result (data engineering), and then creating something that can reproduce acquiring all that data in an automated fashion such that it gets data that's sufficiently the same to produce the same result each time the coin is presented to the system (contraption engineering). Having "the AI grade over the expert's shoulder" needs to be put in more concrete terms, and that's the big challenge here.
Keeper of the VAM Catalog • Professional Coin Imaging • Prime Number Set • World Coins in Early America • British Trade Dollars • Variety Attribution
My EAC Grading Guide is at home so I cannot refute your post with actual quoted from that book!
Perhaps, yo can help me. I have a PCGS Photograde XF 1794 large cent.
Obverse: Rim bruise 2:00. Small, deep rim nicks 3, 8, and 11:00. Small dark green corrosion lump on the "L." Several /4 inch scratches on the bust Eleven scattered, bag mark-looking hits on bust and in fields.
Reverse; same sort of stuff including a little weakness at the bottom.
I contend that six of the best copper dealers/graders in the country would not "hit" the same EAC grade (price for sale). That's why it is complicated. I'll also bet that any EAC dealer, TPGS grader, knowledgeable collector would get the coins "design condition" correct most of the time BEFORE the copper guys "Net" it down.
IMO, it ain't going to happen the way you posted! What I can believe is that one day in the far future AI will help human graders be more precise with their opinions. Thus cutting down on disagreements in house and in the market.
I wasn't saying EAC isn't complicated...I was saying everything is complicated...11 MS grades x 3 letters is complicated, without arguing with Bill Noyes how much a 1/4" ding drops a grade vs. an 1/8" ding.
I guess some of us have a different understanding of "complicated." For me, those 11 MS grades are the easiest to teach!
In my experience, examining a coin (with no time limit and using the proper tools and lighting) and seeing EVERYTHING there is to see on it is EXTREMELY EASY and not complicated at all! It can be successfully taught to any one with an eye-for-detail. That is the BASIS for all coin grading - the first step - properly examining a coin. It provides a coin's "true" technical condition of preservation from the time it was struck. I think many of us can do that if asked. As I posted above: "I'll also bet that any EAC dealer, TPGS grader, knowledgeable collector would get the coins "design condition" correct most of the time..."
Everything after that OBJECTIVE, PRECISE, exam introduces each of the subjective differences that cause virtually all of the problems/disagreements regarding a coin's "acceptable, 80% correct, commercial market grade." These SUBJECTIVE, IMPRECISE factors include:
1.Value
2. Eye appeal
3. Net Grading
4. Ownership
5. Skill/knowledge of the grader of the series, country, type, etc
Good Luck folks, it is what it is!
Insider2 - could you expand on "Ownership"?
He's probably referring to the adage, "Ownership adds a point." He did say subjective and imprecise, after all.
Keeper of the VAM Catalog • Professional Coin Imaging • Prime Number Set • World Coins in Early America • British Trade Dollars • Variety Attribution
To pivot the subject a little bit, would a debate or panel discussion about the future of AI in coin grading be a presentation people would go to at a show, or would it just draw an audience of about 6, like many presentations do? For a debate, I would envision qualified people able to take both sides of the issue drawing their side out of a hat, not so much to convince people of what will happen and "win" the debate, but to lay out all the things to think about and provide 40 minutes of entertainment.
Keeper of the VAM Catalog • Professional Coin Imaging • Prime Number Set • World Coins in Early America • British Trade Dollars • Variety Attribution
An assigned quantity representative of the detracting significance of each mark is then calculated by adjusting the measured surface area of the mark by a factor representative of the relative grading importance of the area on the coin where the mark is located.
Relative grading importance. Kind of a subjective thing is it not?
Whoever is careless with the truth in small matters cannot be trusted with important matters.
Not if it is pre-defined for the series. He's just saying that a hit on the cheek is more important than the same hit near the rim, or something like that.
Folks say that the owner of a coin tends to "push" its grade higher than most others might say.
I like this: "An assigned quantity representative of the detracting significance of each mark is then calculated by adjusting the measured surface area of the mark [and its depth] by a factor representative of the relative grading importance of the area on the coin where the mark is located."
Good explanation of what humans graders do.
Aha - was wondering if you were speaking as a professional grader or collector.
I know conflicts occasionally arise between the TPG's and owners of well known coins/collections and was hoping you'd have some inside tid-bits for us......
This is a topic I am interested in.
The insights I gain here is helping with a new project on digital coin surface evaluation.
Please keep the descriptors and definitions coming!
HJP
Good explanation of what humans graders do.
And its humans that collect coins, not computers.
Whoever is careless with the truth in small matters cannot be trusted with important matters.
I cannot speak for professional graders. I've been allowed to use a stereo microscope to examine any coin I was grading at every TPGS I've worked at! I considered myself a "firewall" to detect counterfeits, repairs, and surface alterations. Ninty-nine percent of the time, I remove a coin from the box, get a naked-eye, gut reaction to its grade and authenticity with incandescent light, pop it under the scope using florescent light, and go from there. By then, most coins are graded and authenticated. Then I use a 7X hand lens under incandescent light to see what other graders/dealers/collectors might think (commercial grade). Then I grade the coin and leave notes. I've learned how some of the top graders see coins and have trained myself to closely match company standards. I saw some advice in one grading room: "Monkey see, monkey do, monkey get rich."
I don't deal with owners of well-known coins/collections. The majority of conflicts I've seen occurred long ago when big submitters complained about the grades on an entire order and threatened to take their business to another TPGS. I believe over the decades this has resulted in much of the gradflation that has taken place!
I do get to handle most of the customer relations that arise from grading or authentication questions. I get to play the devil's advocate and argue on behalf of our customers when they have a legitimate question. That's almost as much fun as looking at coins all day.
Just like the MLB shouldn't get rid of umpires and use software/computer to call balls/strikes. It's part of the charm of the baseball and it helps keep fans and conspiracy theorists riled up and watching. Not much different that PCGS graders rendering opinions on coins. I'm sure each grader has their pet peeves and likes/dislikes that affect their opinion. That keeps submitters scared and excited to get their grades back. I'd think computer grading would reduce the number of coins sent in for grading which is not something PCGS would want to do.
I don't think that toning should affect the grade of a coin. It can affect the price of a coin, but not the grade IMO. People will say toning should affect the grade due to "increased/decreased" eye appeal, but that is subjective and dependent on the collector.
I don't think that a TPG would have to or would want to tell anyone who/what graded something. I'm sure graders wouldn't mind if a TPG start using AI to grade monster boxes of ASE's...
Interesting one. I think that some folks are underestimating the complexity, but it of course can be done. Is it worth doing? Not so sure. I've worked with computer vision, and it is definitely not trivial to detect marks on the huge variety of surfaces that coins present. The amount of programmer hours (which, like grader hours, are not cheap) required for this would be substantial. And the system would still need to be supervised for quite a while. Fortunes can be made and lost on grading points.
I think starting with moderns like (untoned) ASEs would definitely ease some of the difficulty. For classic coins, it could be super challenging, especially early era US where there is significant die to die variation. Computers don't see like we do, they are precise. If anything is out of place, a die break, etc, it can throw things off.
As a personal project, I once did some vision recognition on coin images. Even consistently getting orientation of the coin was challenging! And I was working in a single series.
Aercus Numismatics - Certified coins for sale