AI Generated Album Pages - Curious On Feedback And Thoughts

I asked AI to generate album pages for some argentina stamps I have and this is what it came up with.

It would be in a layflat binder in sheet protectors so hole punches are not necessary.

Left side is text, right side only stamps.

Curious what you guys think.

Thanks for posting this, just recently I have been wondering how AI can help with stamp collecting and generating album pages seems like a great use case. The new models released just over the past month (Opus 5.5 and Astra 6) are just mind blowing for what I use them for, but haven't tried anything philatelic yet.

Can you share what prompt and what model you used to generate it? Did you have to give it details about what stamps to include and what the dimensions and characteristics were or was it able to find that on its own?

I think the pages look great. Have you tried printing them and filling with stamps? I would love to see finished result.

Probably not what you want to hear, but this raises two questions in my mind:
1. In my experience teaching history, AI repeatedly makes mistakes, often covering them up with fatuous generalizations that are misleading and sometimes just getting facts incorrect. What's to say any write-up you end up with won't have similar errors?
2. "Write-ups" are written up by the collector themself whom we depend on to have checked the facts and claims as accurate or what is the point of having them? Using AI circumvents that both lazily and perhaps misleadingly. I'd be very wary of any stamp presentation written by AI which should, of course, be labeled as such.

And I'll add a third concern. Stamp collecting in its more detailed manner like yours is about doing the research and evaluating its accuracy before presenting it. It relies on the collector making the effort. Is that what's being done here -- or is this simply printing out pages you hope are accurate? Would this be acceptable in a stamp exhibition? I would think not. I pretty much don't really care what any individual collector does with their collection for themselves, but when it comes to presenting history and facts it ought to be correct and that means checking for accuracy about both the philatelic facts and the historical facts, don't you think?

It's worth discussing as AI rocks everyone's boat these days.

"I asked AI to.." is not going to get you very far beyond the initial excitement.

Have you researched how to support creative work with AI?

For one, you need to develop and maintain a "skills" file specifying how the LLM should create the pages. Otherwise, the pages will not be consistent spanning sessions, models etc.

Create the skills file, post results, then people can discuss.

Regarding "AI write-ups": If you want to fill your album pages with that it's up to you. For me, it's a no-go.

I'll add my two things.

AI is a crutch wherein you can present a complex topic and have zero understanding of that topic because you did little to no work assembling the information. It really is glorified plagiarism.

AI vacuums all of what it gives you from other peoples work. Have you ever seen AI acknowledge attributions or copyrights? If it exists digitally AI "borrows" it.

I'm waiting for some intellectual property theft cases to be brought vis-a-vis AI.

A brief survey of an SG catalogue reveals that Belgrano appears on stamps in 1867 and 1892 and then maybe nothing until this 1974 issue, just as Peron was shuffling off this mortal coil. Otherwise it's mostly Rivadavia and San Martin. The cruiser 17 de Octubre was renamed the General Belgrano subsequent to a 1955 coup which overthrew Peron. The date was significant for the Peronists, hence the change.

[edit; aircraft carrier->cruiser ]

Tsmatx: Astra 6. I prompted it to create a page based off a known catalog number I had. It found a public online specialist catalog itself and used that information to create spots for the varieties listed. It initially did not know sizes, but now pulls that information automatically from other online sources where that information is documented . It flags information it does not have and leaves that information blank and does not utilize it, hence why there is no boxes for the mounts for the margin stamps as it is unsure of the size

DrewM: You bring up some great points. Hallucination has been a problem with AI for a while. That said with the new models recently released, it has been drastically improved. These aren't the free/cheap AI your students have used in the past. In fact, it has the ability to provide primary/secondary source citations for all claims.

I guess for someone like me, I don't wish to dedicate years of my life to the study of a handful of random topics that won't generate growth for me, this is the acceptable alternative. If the information is verified correct, what is the problem?

This ties into Rogdcams excellent points as well. It all depends on whether you view information as free and available or restricted behind a wall of study/academia and time investment.

ClassicalStamps: Yes I am aware of skills. Would you be willing to expand on why ai writeups are a no-go for you?

And to everyone, what things for you specifically make a great album page?

Astra 6 really is a game changer. I use it for work and am amazed every hour what it is capable of, but I can't use it for anything fun like stamps.

For me personally, I prefer more minimal pages (like Steiner, or Scott) - I don't need the long writeups and graphics etc. I believe AI would be very good at producing minimal pages just as well as 'fancy' pages like yours. It is very good at page layout. I would also expect it to be very, very good at producing things like AlbumEasy layouts. I'd bet you could feed it an image/PDF of a printed album page and it would be able to reverse engineer to an AlbumEasy layout, which would be very useful.

I don't see anybody pointing out any specific errors in the actual pages, only claiming that there MIGHT be something wrong. But you may get more concrete feedback if you tried creating pages for more widely collected stamps (like US 19th century) since more people are familiar with those stamps and there would be more to critique.

The only 'flaw' which I see in the pages is the reference to a certain online stamp catalog. I understand why that is (because it is the only one freely accessible), and I'm sure you could prompt it to avoid mentioning it. It would look more professional and polished if it referred to Scott or another major catalog (which, of course, has hurdles because those catalogs are unfortunately not free).

The perspective might depend on whether you created the information AI is stealing or you are benefiting from others work.