Winds Of Change In The Philatelic World

Crack open any of the Philatelic Foundation "Opinions" books and it becomes clear that flesh and blood are required to expertize.

gvol21 and eyesonwall raised crucial questions. Spent some morning coffee time looking at the PF, PSE, and APS sites for mention of opportunities to be an expertizer and found only this from APS:
https://stamps.org/news/c/news/cat/...-expertizing

Who could possibly fill the shoes of someone like Frank Mandel (an expert's expert). Or those of Eric Jackson, Richard Friedberg, and Ron Lesher (revenues of all kinds). Jim Lee is semi-retired (Civil War era postal history, essays & proofs). Dealers are in a better position to become expertizers than collectors, tho veteran specialty collectors could potentially fill many niche expertizing vacancies. Don't know what it would take for expertizing firms to "recruit" the next gen of expertizers, but the points raised here expose a real need IMO.

Don, sure - in the same way they use Google right now to look stuff up. AI will become a tool like a search engine or anything else. Computers or AI or whatever don't expertise stamps; humans do.




A discussion of these topics is what I was hoping this thread would elicit.

Actually not automatically true. Many of the best expertisers have been and still are past or current auction describers who were also collectors for many years rather than dealers.

Are one flavor of "stamp dealer," like it or not.

Edit to add:




Of course when AI IDs all of the forgeries what is left in a group of stamps ARE NOT GENUINE stamps, BUT POSSIBLY GENUINE stamps.

For example take US 1203 and 1204. Will AI be trained to find original yellow inverts of 1203 and separate them from the purposefully yellow inverted reprints, Scott 1204? Currently there are no experts who can do so.

Merely identifying all currently known and verified yellow inverted 1203 stamps and covers does not produce the ability to find other genuine but not identified 1203 yellow inverts.

As they were different press runs and thus different press inking occasions, a full spectroscopy analysis of all existing known 1203 with yellow inverted against a large sample of known 1204 stamps may find a difference. But first the yellow inverted 1203 would need to be compared with normal 1203 specimens to see that the ink was consistent within the original 1203 runs. The "known" 1204 would likely need to be taken from a not yet opened post office pad of the issue to verify they are in fact the 1204 with out possible genuine 1203 yellow invert.

Now who will undertake the cost to get this data in a clean form to feed into the machine learning and at what price will it be? Will it be offset by finding valuable new 1203 inverted yellow example and with each find won't the "valuable" index decrease? I think not. But of course AI could be trained to check the paper type, gun type and amount of wear on the perf pins used to perforate.

Here a question I would like to see discussed in this topic, if AI can be trained to find genuine stamps (not speaking to just 1203/1204) will it be able to learn how to produce reproductions which can pass for genuine examples? If so, then that would be great, everyone could fill all the spaces in their album, right?

I would disagree. "Dealers" are direct sellers. "Describers" are just that, people who give a written explanation of an item. They help, but they have no direct control over whether said item sells or not. I have known several describers; I have never heard any call themselves dealers. Unless they were doing both they would not.

We have some "experts" of our own on this Board.

For example, see the Show Your US 1851-57 Imperforate Stamps at...

https://goscf.com/t/72775&whichpage=184

With a knowledge base represented by such a group, and the published resources, could AI assist in the plating of stamps from this issue?

Would that be legitimate application, and would it be a useful thing, or simply remove much the fun from collecting the area?

No doubt about that, and covering many different areas and countries.

Perhaps we should ask the experts here if they have ever formally expertized and if so why did they stop or are they still doing it. If they have never employed their knowledge and skill to formally expertize it would be educational to know the reasons why. If there is a shortage of expertizers the first step in solving the problem is understanding why.

Expertisation, much like many other things in philately today, often devolves to those least able to do it. Two simple reasons, time and money. Most of the individuals who possess the requisite knowledge to expertise stamps and covers with 95 to 98% accuracy can make more money doing other things and have less time to spend doing "charity" work with little, or no, compensation.

Good point Richard.
For the things outlined in Richard's post above, I see the future of expertisation moving towards AI for these reasons;

Automation of Routine Tasks: AI can automate repetitive and routine tasks, allowing experts to focus on more complex and creative aspects of their work. For example, AI can help with identifying more common stamp varieties, which frees them up to concentrate on more challenging patients.

Enhancing Expertise: AI can act as a tool to augment the capabilities of experts**. It can analyze vast amounts of data quickly and provide insights that experts might otherwise miss. This can make experts more efficient and effective in their decision-making. Anyone here think that getting a cert back quicker is not a good thing?

Reducing Error Rates: AI can help reduce human error. This is particularly crucial in expertizing areas such as plating where precision is essential.

Creating New Roles: The development and deployment of AI can lead to the creation of new roles in expertisation. Experts may need to adapt and learn to work alongside AI systems, maintaining oversight, and ensuring the ethical and responsible use of AI.

AI has the potential to assist the hobby IF the trend is a reduction in human experts (which I do not agree with) over time.

Edit: If AI was handling the more common expertisation tasks, would expertisation be more attractive to those without the time now? Or those who do not want to invest significant amounts of their time on the more common material?
Don

**I hope folks stop looking at AI as simply a 'chatbot'. 'Chatbots' like ChatGPT or Bard are currently prominent applications in the realm of natural language processing (NLP) but there are unlimited other ways in which AI can (and will) change philately.

Seems as if the common denominator that is needed to implement AI in a meaningful, serious way and bring human's onboard is money and lot's of it. Lots of it at the top of the hobby food chain. Will someone step up?

I assume that this will unfold with philately being (far?) behind the curve and this will let others deal with the majority of the risks and costs associated with early adoption. I think our hobby will adopt once the tech/apps have been in place for quite some time.
Don

Edit: I apologize to all for beating the 'AI drum'. I am fine with philately being behind the bleeding edge tech curve but feel that being 2 decades behind the curve is not healthy. I think we (the hobby) really missed the timely adoption of the internet curve and hope we can do better with AI.

Don,

That would be the prudent thing to do. As we are discovering in academia, the use of AI is fraught with intellectual property issues and legal implications.

Dan,
Agreed, but please see my edit above that I was writing at same time you posted.
Don

Don,

I agree... there's conservative and risk-averse, and then there's "2023 Amos Media poised to leap into the 1990s" levels of conservative.