A demonstration of using AI to answer a Capped Bust Half Dollar question

Below is the summary of a question I posed to ChatGPT today. Some of you may find it of interest, so I thought I would share.

I am trying to find a detailed comparison between the various designs used for specific years in Capped Bust Half Dollars. I am a Type collector, so design changes are important to me. I am NOT a variety collector, or a specialist in any one coin series.

Within the Type 1, Lettered Edge (1807-1836) Capped Bust Half Dollars, I heard rumors that there is actually two design types in that time period. If so, I may want to include it in my 8080 Type set, which is an expanded 7070 Type set.

Normally, I turn to PCGS Coinfacts, which usually has a nice little graphic showing mid-year design changes. Now this design change also coincided with a date size and letter size change, so all the graphics and answers seem to focus on the comparison between small vs. large sizing, and miss completely that the design itself was updated.

When I could not easily find a photographic comparison on PCGS Coinfacts related to my Capped Bust Half Dollar question, I decided on a whim to try Machine Learning / Artificial Intelligence engine, specifically ChatGPT 5.2.

My past experience with AI has definitely been a mixed bag. My overall assessment as of 2025 was that AI gives reasonable answers that sound plausible to a non-expert, but that if you have subject matter expertise, the answers are either partially wrong or misunderstand the nuisances.

I am NOT an expert in the Capped Bust series, so to me the following AI answer sounds plausible, and I am sharing it here for two reasons.

• I want to show how AI engines might be useful to find answers to questions related to our hobby.
• I want to know if this answer is correct. Fully correct, partially correct, plain wrong, or just utter gibberish /nonsense. (I've seen answers like this)

My Question posed to ChatGPT:

Answer:

My initial reaction was that I found this answer to be very complete. But what I really want is a visual overlay or a side-by-side comparison. So, next I will ask AI to produce such an image.

Image Produced:



I have absolutely no idea why it used a super low-rez / fuzzy image, so I also asked it to change that. May be due to privacy or copyright issues with images. My follow up prompts to fix the resolution did not bear fruit.

Even where AI placed the arrows seems a bit odd, for example, look a the "chin / jaw" arrows. Pointing to the nose?

I am thinking the answer may very well be correct, but certainly the graphic is deficient. Overall, I did find some detailed answers that I was having a hard time finding with general googling.

I am VERY interested to here from Capped Bust experts. How did AI do? Is there a better public source for photos and answers related to this updated 1834 Obverse hub?

I hope you find something interesting or helpful about this post.

 

Well Matt, you get full points from me for this exercise in and of itself plus all the thought that went into it.

And yeah, you could call that a Like

 

I spent some time in Adobe Photoshop overlaying the 1833 over the 1835 (before and after the Obverse Hub was changed):

From that I can see, really the bust area, chin, and nose have minor changes. I guess this is why very few people break them into distinct Types.

Still, I could see someone doing it. (like me). The CAC registry actually considers these distinct, so if anyone is competing over there for a Type set, they would need a (1807-1808), (1809-1834), and a separate coin for (1834-1836) which they refer to as "Slightly Modified Portrait".




Of course, it would be perfectly reasonable to keep all of 1809-1836 as a single "type" since the changes really are fairly minor. I didn't know there was a Hub update, and hence my journey in the OP to find answers. Now that I see how minor they are, I am hesitating on making the distinction.