• When you click on links to various merchants on this site and make a purchase, this can result in this site earning a commission. Affiliate programs and affiliations include, but are not limited to, the eBay Partner Network.

707comics

Member
  • Posts

    1,174
  • Joined

  • Last visited

Posts posted by 707comics

  1. On 8/21/2023 at 1:25 AM, Get Marwood & I said:

    Right, thanks. If correct, your process timing calculations strengthen my theory that AI might be grading a lot of the books. Maybe the human graders just do the pre-moderns which seem to represent a much smaller part of the overall volume if BCs extracted data is correct. If you were running a grading company, and got a human to count the pages, do a resto check and look for internal issues on every submission, and the statistics then showed that 99.8% of moderns always had the right number of pages, no resto and no internal structural issues, what would you do? Carry on doing that, or accept the risk of that tiny percentage of books with issues getting through and then being discovered?

    Carney's AI might be sorting the books into work queues and sending everything before a given date to the humans for grading and grading the rest itself - probably at the same time - based just on the covers. It then directs those to encapsulation having populated the records. AI may even then do some QC. If it can identify any title from a possibility list of millions, it could match the cover title to the label etc. There are lots of opportunities for it to add value, I'm sure.

    Anyway, it's not our business what CGC do or how they do it. It's fun to speculate, but all we can do is decide whether to pay for the end service. If we're happy, we carry on submitting. If we're not, we have the option of not using them (although most seem to carry on if this thread is any indicator). When the process works, the end product looks fabulous. When it doesn't, it gets posted about here and in other places online. Mike says that CGC has a very low failure rate. That's probably true. When greggy does one of his 98 thousand graded book acquisition posts they all look fine to me. So maybe this thread and others do only represent a very small failure rate. Thousands of happy people, who say nothing, scores of unhappy people, who post here. Who knows. Those kinds of figures would be acceptable in most industries though, I'll bet. 

    I've only joined in here because, once again, I've come to the forum looking for fun and interaction and this thread stands out. CGC don't pay attention to it. They don't change their practices or anything because of it or any other thread. They have their model and it's working for them. Use it, don't use it. If you complain, Mike may send examples of systematic failure through, and then nothing will happen. Five years in, and the solution to Newton Rings - after all that debate - is that they can't / won't fix them and they're now normal. For me, that is why I pause before submitting. It's not because I fear being in the tiny failure rate group (although I'm lucky like that). It's because if the failure rate is very low, then CGC should be remediating customers. The cost of doing so would be low, as the failure rate is low. But they're not. They're trying to normalise failure and have you pay for it. That's the reason I'm hesitant to use them.

    I asked this question a while back (i.e does CGC use any type of A.I. in the grading process) but never got a concrete answer. 

    IMO, It is the only way to really scale rapidly and in a cost effective way. As the volume increases and the demand for "humans" goes up to keep T.A.T.s somewhat reasonable, consistency becomes a big issue. Come to think of it, so do cost and talent availability (only so many graders / those interested in becoming one, training time, training cosr, etc.). With A.I. books can be graded anyplace, anytime. The training process is always ongoing and scaling is a much different scenario. Instead of being bound by how many human graders you have, it quickly becomes how many photos can you provide the A.I. with a day to score, etc. I can see how Computer Vision could also be beneficial in the Q.A. procedure. It could quickly check centering in the encapsulation, maybe look for newton rings (hard to see on camera though), labeling issues, etc.

    There are all kinds of arguments for and against A.I. that are probably more geared for the water cooler but speaking strictly from a business perspective I can see how an operation like CGC would be investing heavily in A.I. and computer vision to streamline / scale their operations. I would love to hear more about if / what / thoughts on AI / CV from the CGC team but not sure how likely that is to happen.

    2c (FWIW, I just submitted another batch of books last week)

  2. On 8/7/2023 at 1:26 PM, Robot Man said:

    Oh, no Ray passed away? Hadn’t heard. That is very sad. I knew Ray for many years. Always an adventure picking his boxes and talking comics. I remember buying a pile of press proof covers when he turned them up. A great guy. 

    I had not heard myself until the just the other day. I was really saddened to hear, he was such a great guy. I always enjoyed chatting with him at the shows and looked forward to seeing what kind of esoteric goodness I could find buried in his boxes to bring home with me. He always had some book that seemed to call to me :)

    I found so many cool books in his boxes and he had a story for each one (if it wasn't about the book it was about the story or artist inside). @Robot Man I think I have a few of the Atlas war and western proofs you are referring to. I was not collecting when he was selling them originally but I acquired a few here on the boards later (either from you or @catrick339 if I had to guess). My first copy of Atlas War #11 was from one of his boxes amongst many other 1950s Atlas war and GGA titles. 

    Always sucks when one of the good ones leaves us.