AI For Card Grading?

[discodanman45]

I wonder if DraftKings is ok with them taking additional investments. How much did DraftKings invest in them? A free t-shirt to give away at Fanatics Fest?

[Snowcat]

Fyre Fest vibes.

Can't wait for the Netflix doc

[ScattMan]

They deleted their twitter account and posted this on their website

[discodanman45]

Looks like there is a Kyle Cook that fits the Amazon part of the description. He deleted his Linkedin account, which is pretty strange. The picture matches up with Bea, whose name is on the LLC.

https://aws.amazon.com/blogs/publics...n-data-on-aws/

This is very, very strange and I don't know what is truth or made up nonsense at this point. Creating a LLC through the company they did is extremely amateurish and not professional at all, it is what a tiny t-shirt print store would do or a scamming company. Their slab and photoshop skills are what a 5 year-old could do, not what you expect from someone that seems intelligent from their Amazon report. I am at a loss for what is going on.

[BoSoxFan1999]

Originally Posted by Scottish Punk

I think you are right scattman, this smells of a scam.

The more I learn, the more I think of the Monorail episode from the Simpsons.

[salsdali]

the fact that there's so much talk in this thread about a new grading company shows how unsatisfied customers are with the current situation.

If PSA was so great no one would even be talking about this new startup.

PSA should be shaking in it's boots if a new scammy company can command attention from it.

[rfgilles]

Originally Posted by salsdali

the fact that there's so much talk in this thread about a new grading company shows how unsatisfied customers are with the current situation.

If PSA was so great no one would even be talking about this new startup.

PSA should be shaking in it's boots if a new scammy company can command attention from it.

The only attention they are getting is because it is so scammy. It is the BO forum smelling blood.

[SteelPen2]

Originally Posted by salsdali

I think the main flaw is that all scanners are set up to scan in 2D. Surface imperfections are actually a 3D shape so really the card (or earthenware) would have to be scanned with a 3D scanner even though it seems like a 2D "flat" surface.

3D scanners have come way down in price so I don't see why this wouldn't be a feasible method. The hard part would be quantifying surface imperfections to a grade.

I agree.

[CorndogWasp]

How does AI rotate the cars to check the surface? And then flip the card and examine the back? Does this AI come with this machine?

[bojesphob]

Originally Posted by CorndogWasp

How does AI rotate the cars to check the surface? And then flip the card and examine the back? Does this AI come with this machine?

I would imagine it really depends on how the AI is written. Could be they do a 3D scan. They could also do super high resolution 2D scans that capture the surface, corners and edges, and centering. That said, the process of getting the images or scans is separate from the AI grading process, as AI isn't about hardware or software control in this type of scenario, it's about using information (in this case images) to make classification determinations.

[BoSoxFan1999]

I wonder if their "CEO" was even at Fanatics Fest.

[CorndogWasp]

Originally Posted by bojesphob

I would imagine it really depends on how the AI is written. Could be they do a 3D scan. They could also do super high resolution 2D scans that capture the surface, corners and edges, and centering. That said, the process of getting the images or scans is separate from the AI grading process, as AI isn't about hardware or software control in this type of scenario, it's about using information (in this case images) to make classification determinations.

Well while the worker is physically handling the card doing all these steps they can just look at the card for 30 seconds and grade it. Done.

[anusinha]

Originally Posted by BoSoxFan1999

The more I learn, the more I think of the Monorail episode from the Simpsons.

Is there a chance the slabs will bend?

[bojesphob]

Originally Posted by CorndogWasp

Well while the worker is physically handling the card doing all these steps they can just look at the card for 30 seconds and grade it. Done.

That's definitely what they do now. Although, once the scans/pictures are done, if the trained model is sufficiently data populated, the process to grade would be several thousand times faster than doing it manually. The longest part of that process would be getting the scans/pics, and they could do thousands of grades in seconds at one time, possibly even 10s of thousands in that same amount of time one person could do one grade. I sort of liken it to the database work I do: I could go through and manually update records in the database for person information, but I'd have to go through and compare each record to the source. Each one wouldn't take more than 10-15 seconds. But, there are over 1 million records in our person information for students and employees, and the database can do the comparison and update in less than 3 seconds for all of them. If the classification in the AI model is sufficient, then the computer can do it FAR faster than any human could. The process to prep it to be able to be graded, though, I don't think can be automated easily, and that would then be the largest time sink. But, that is assuming that the grading can be done consistently, reliably and accurately, which I still (personally) have my doubts about considering all of the potential variables involved.

[CorndogWasp]

How? you still have to physically stand there and move the card around and flip the card around. Do they have a machine for this part? Also taking from the pile, grading it and putting it the new pile also requires a humans.

I think its easier to just have them look at it since they are already holding it anyway to move it around to the different AI machines. All this AI stuff is being described as something that takes much longer than a person grabbing the card and looking at it for seconds.

[rfgilles]

Originally Posted by CorndogWasp

How? you still have to physically stand there and move the card around and flip the card around. Do they have a machine for this part? Also taking from the pile, grading it and putting it the new pile also requires a humans.

I think its easier to just have them look at it since they are already holding it anyway to move it around to the different AI machines. All this AI stuff is being described as something that takes much longer than a person grabbing the card and looking at it for seconds.

How do you know someone has to flip the card? You really think someone hasn't figured this out. Have you heard of TAG and AGS?

A human can look at the card, but how do you know they will grade it consistently and accurately.

[salthill]

Originally Posted by Gary

Anything can be done if you throw enough money at it,both PSA and CGC have deep enough pockets to get it done.that assumes they have the will to do it.So far that seems to be a no on the will part.im sure both PSA and CGC have done cost to profit ratio.must not be worth the employee savings at this point.it will happen someday same as the fast food kiosks happens.when it makes sense profit wise someone will do it.

PSA already bought an automated grading company (Genamint) in 2021, and made its founder PSA’s VP of Product Management.

[YeahBuddy]

They both have openly admitted to at the least using AI to help assist in different parts of the process. Even if its simply helping identify.

[bojesphob]

Originally Posted by CorndogWasp

How? you still have to physically stand there and move the card around and flip the card around. Do they have a machine for this part? Also taking from the pile, grading it and putting it the new pile also requires a humans.

I think its easier to just have them look at it since they are already holding it anyway to move it around to the different AI machines. All this AI stuff is being described as something that takes much longer than a person grabbing the card and looking at it for seconds.

A 3D scan or photos would only a few seconds per side. Less if they have some sort of conveyor system that does multiple cards in a batch in succession. Grading manually is more than just looking at it for 30 seconds and giving a number - it's inspecting the edges under magnification, corners under magnification, then going over the surface on both sides. All of that takes time. It definitely won't be manual work free, for sure, but AI being able to do the actual grading will cut times down on that step a lot.

[CorndogWasp]

Conveyer system? Sounds expensive to invent. They have an AI machine that uses magnification and can change the lighting in ways to see the surface?

Have any of these machines been invented yet? Or is this all “loose” definitions for the word AI? Call already existing technology AI after the fact?

[bojesphob]

Originally Posted by CorndogWasp

Conveyer system? Sounds expensive to invent. They have an AI machine that uses magnification and can change the lighting in ways to see the surface?

Have any of these machines been invented yet? Or is this all “loose” definitions for the word AI? Call already existing technology AI after the fact?

AI isn't "hardware" by any stretch, so those would be things that the grading company would need to determine outside of the AI stuff. AI is software in the computer that does analysis, and in this instance classification. It takes inputs, such as text for things like ChatGTP, or images like the image generators, and then processes the data to come up with the end result. How the images are obtained would definitely be a part of the process for figuring out a solution for this, but it's a different piece of the puzzle. AI only cares about data inputs and outputs, and what it can do is entirely dependent on the data that's available to it, and how the code is written to generate the classification models. The better the scans/pictures, and the more it can compare against, the more accurate it's going to be when determining a grade.

We just had this conversation at work yesterday at an in person meeting for the department. AI is touted as this magical thing that can do all this super incredible stuff with little to no effort and data, but that's all marketing fluff.

Another way to look at it is the old IT adage: Garbage In/Garbage Out. AI is only as good as the data it has to train the models off of. The better the scans/pictures, the better the classification results. If the data to train the models is terrible, or it can't easily distinguish condition issues, then its usefulness is going to be extremely limited.

[rfgilles]

Originally Posted by CorndogWasp

Conveyer system? Sounds expensive to invent. They have an AI machine that uses magnification and can change the lighting in ways to see the surface?

Have any of these machines been invented yet? Or is this all “loose” definitions for the word AI? Call already existing technology AI after the fact?

Yes....

https://taggrading.com/pages/dig

[Blake050]

Originally Posted by anusinha

Is there a chance the slabs will bend?

Not on your life, my Hindu friend!