Answering numismatic mysteries with ChatGPT!

Zoins

Extending the use of AI in numismatics from new coin designs, I asked ChatGPT to pretend it was a coin and tell us about itself in the Do you wish that a coin could tell the tale?.

This got me wondering what else could ChatGPT tell us about coins.

Here are answers to the locations of coins in my answer to the Do you believe mega-ultra rarities are still out there, and entirely unknown? thread.

I wonder where it's getting it's information for these, including the gold Libertas Americana at the ANS and the Bibliothèque nationale de France in Paris, as well as the 1964 proof Peace dollars.

Question: Where are the original two gold Libertas Americana medals?

Answer:

The original two gold Libertas Americana medals are housed in the collections of the American Numismatic Society in New York City and the Bibliothèque nationale de France in Paris.

The Libertas Americana medal was created in 1783 by Benjamin Franklin to commemorate the role of France in the American Revolution. The medal features the image of Lady Liberty holding a pole with a Phrygian cap, a symbol of liberty, and an olive branch, representing peace. The reverse side depicts a bust of Benjamin Franklin and the date of the signing of the Treaty of Paris, which ended the war.

While the original dies used to strike the gold medals were sent to France to produce additional medals, only two gold medals were struck from the original dies. One was presented to King Louis XVI of France, and the other to the President of the Continental Congress, Thomas Mifflin. Both of these original gold medals have since been lost to history, and the surviving examples are all struck from later dies.

Question: Where are the original 1964 Peace Dollars today?

Answer:

The original 1964 Peace Dollars were never released for circulation and were melted down by the US Mint.

The Peace Dollar was first issued in 1921 to commemorate the end of World War I and to symbolize peace. The design, which featured the image of Lady Liberty on the obverse and an eagle perched on a branch on the reverse, was created by sculptor Anthony de Francisci.

The Peace Dollar was minted from 1921 to 1928, and then again in 1934 and 1935. However, production of the coin was suspended in 1935 due to a glut of silver on the market and the need for the US Mint to focus on producing circulating coinage during the Great Depression and World War II.

In 1964, a small number of Peace Dollars were struck at the request of coin collectors, but these were all produced as special proof coins and were never intended for circulation. The coins were struck in small numbers and were sold to collectors by the US Mint. Any remaining unsold coins were melted down.

So, to answer your question, the original 1964 Peace Dollars were melted down, and no known examples of these coins exist outside of a few rare examples that were struck as proofs and sold to collectors.

yosclimber

https://nymag.com/intelligencer/article/ai-artificial-intelligence-chatbots-emily-m-bender.html

291fifth

What a great way to spread misinformation. The dangers associated with AI are going to come into focus very quickly in the near future.

dcarr

I have never paid much attention to the AI ChatGPT before. Reading the post above it appears that it finds bits of information (or misinformation) on the internet and splices it together. I see no evidence of it thinking for itself, or any indication that it is aware that some information out there might be false. A true AI system would report only what it actually knows from experience to be true, and beyond that, "I don't know" would be the most reasonable reply.

Exbrit

Main concern - garbage in garbage out and bias. Other than that - who knows…

ricko

We have much to learn about AI...and the inherent dangers it poses. Right now, I fear the greatest danger is to education, the downfall of which will kill creativity and eventually doom the human race to servitude to power. Cheers, RickO

jacrispies

I played with ChatGPT and asked some specialized coin questions. It responded incorrectly with confidence. The program is useless to me now, knowing it passes false information as true.

291fifth

@jacrispies said:
I played with ChatGPT and asked some specialized coin questions. It responded incorrectly with confidence. The program is useless to me now, knowing it passes false information as true.

But wait ... since you read it on the internet you know it has to be true!

Shane6596

Tramp

Just Amazon Alexa on dope!

At least when Alexa didn't have an answer it responded that it didn't have an answer or didn't understand the question. Now they have something that will give a lie and present it as fact. Misinformation, propaganda or what ever you want to label it.

tyler267

Chatgpt is really good for writing letters memos procedures ect, super efficient. Problem is it gets the a lot of things wrong. I use it almost everyday but I research the facts. It's only been out since November so it should get more accurate over time.

BStrauss3

@tyler267 said:
Chatgpt is really good for writing letters memos procedures ect, super efficient. Problem is it gets the a lot of things wrong. I use it almost everyday but I research the facts. It's only been out since November so it should get more accurate over time.

I don't think so. They will be releasing new models which have been trained on newer data, but this model was trained on data up through mid-2021. They are working on plugins to search the web for more current data.

Other AI models take different approaches. Bing (Microsoft) is more of a natural language front end to a search and it returns results with web references.

It's really early in the life cycles and we're seeing both solutions in search of a need and struggles to figure out how to make it pay for itself.

If we've learned anything from the voice assistants (Alexa, Siri, Google) it's that people use them for quick questions such as "What's the weather out" but not for the eCommerce functions that were expected to pay the freight.

tyler267

@BStrauss3 said:

@tyler267 said:
Chatgpt is really good for writing letters memos procedures ect, super efficient. Problem is it gets the a lot of things wrong. I use it almost everyday but I research the facts. It's only been out since November so it should get more accurate over time.

I don't think so. They will be releasing new models which have been trained on newer data, but this model was trained on data up through mid-2021. They are working on plugins to search the web for more current data.

Other AI models take different approaches. Bing (Microsoft) is more of a natural language front end to a search and it returns results with web references.

It's really early in the life cycles and we're seeing both solutions in search of a need and struggles to figure out how to make it pay for itself.

If we've learned anything from the voice assistants (Alexa, Siri, Google) it's that people use them for quick questions such as "What's the weather out" but not for the eCommerce functions that were expected to pay the freight.

I will defer to your expertise in the field of artificial intelligence, I'm not an expert or authority in the field. I spend a lot of time writing letters on tax law, I find Chatgpt to be highly efficient in generating rough drafts, but the information needs to be vetted. At this time it is not a replacement for old fashioned research, just a faster way to write. My assumption is that products like Chatgpt will become more accurate over time. I think it's pretty amazing what Chatgpt can do even in it's infancy.

Zoins

@tyler267 said:

@BStrauss3 said:

@tyler267 said:
Chatgpt is really good for writing letters memos procedures ect, super efficient. Problem is it gets the a lot of things wrong. I use it almost everyday but I research the facts. It's only been out since November so it should get more accurate over time.

I don't think so. They will be releasing new models which have been trained on newer data, but this model was trained on data up through mid-2021. They are working on plugins to search the web for more current data.

Other AI models take different approaches. Bing (Microsoft) is more of a natural language front end to a search and it returns results with web references.

It's really early in the life cycles and we're seeing both solutions in search of a need and struggles to figure out how to make it pay for itself.

If we've learned anything from the voice assistants (Alexa, Siri, Google) it's that people use them for quick questions such as "What's the weather out" but not for the eCommerce functions that were expected to pay the freight.

I will defer to your expertise in the field of artificial intelligence, I'm not an expert or authority in the field. I spend a lot of time writing letters on tax law, I find Chatgpt to be highly efficient in generating rough drafts, but the information needs to be vetted. At this time it is not a replacement for old fashioned research, just a faster way to write. My assumption is that products like Chatgpt will become more accurate over time. I think it's pretty amazing what Chatgpt can do even in it's infancy.

That's my experience as well. ChatGPT dramatically reduces the time to have a draft to review, but from my use, the information should still be reviewed for accuracy and context before release.