AI isn't as smart as you think it is

I uploaded this ancient coin image to Chatgpt and asked it "what kind of ancient coin is this". Read the reply below. Can it be more vague.

 

Log in or Sign up to hide this ad.

FAFO?

Given my opinion about stochastic parrots, I'm not surprised at all. If they string two correct phrases together correctly, that's when I get surprised.

 

AI is only as smart as the persons that programmed it.

 

AI is just getting started. When I was in highschool the internet was dial up. This is like the former CEO of Microsoft criticizing the unveiling of the first iphone because it didn't have a keyboard. I don't know that it's specifically being developed to identify random ancient coins yet which for many, there is very little known data on anyway.

 

Hi @-monolith- ,

I assume you knew it was CPE-B310 (Sv-351). The counter marks are a cornucopia and two ears of grain, both in ovals. The coin is overstruck on a Seleucid bronze, probably SC 572 (Antiochos II). For a similar one, overstruck and similarly countermarked, see http://ptolemybronze.com/ptolemy_series.html and scroll to the section “Ptolemy II - Ptolemy III - Ptolemy IV Queen Portraits”.

Take that AI.

- Broucheion

 

Actually that's quite a reasonable reply for a general purpose chatbot.
You really need to test a bot specifically programmed for the specific special purpose subject of ancient coins.

Ross G.

 

Arthur C. Clark foretold this nonsense years ago in a short story. Wholly Crap! It's here....

 

Which one do you have in mind? Wondering if I've missed one...

 

Good Lord, it was so long ago Jeff. I read it in a collection of short stories by him back in the early 70's. The gist of it was that human kind became so dependent upon computers to run everything that the computer began to disrupt everyday life. I distinctly remember him mentioning the disruption of traffic lights by the entity and everyone hearing the laugh and chortle of a young child while the mele erupted.

 

"Dial F for Frankenstein", 1965. Great story. I first read it as a kid sometime in the early 70s.

Not so much that everything was dependent on the phone network (although it was), but that adding so much complexity to it accidentally produced intelligence.

That's a common theme in early stories imagining the birth of AI. With the benefit of hindsight from The Future, it's seeming less plausible these days. Much more likely that some people or companies produce something they don't (and could never) understand, and it takes off in pursuit of its own goals.

 

I tried again with a common Gordian III that has a clear inscription and portraits. Anyone can easily identify this coin in 30 seconds using the OCRE database. Chatgpt took the same amount of time but didn't give accurate information. Surprisingly it didn't use the OCRE database; perhaps copyright issues?

Here is the actual information for the coin:

Obverse: IMP CAES M ANT GORDIANVS AVG
Radiate, draped and cuirassed bust right​

Reverse: CONCORDIA AVG
Concordia, draped, seated left on throne, holding patera in extended right hand and double cornucopiae in left hand​

References:
RIC IV, Part iii, 35​

Authority: Gordian III

Denomination: AR Antoninianus

Mint: Rome (240 AD)​

 

Key is not yet. Give it time.

Look at the computer or device that you are using to view this. Where was this technology 40 years ago?

Technology is evolving at an unbelievable rate.

 

Out of curiosity, I asked Chat GPT about a certain "star and crescent tesserae series." It gave me a detailed description of said series. A detailed description that was directly copied from MY cointalk post about the very series.

Now, even though I am confident in what I wrote, it was mainly speculation about the purpose and themes of said series. It took my speculation and said that it was a concrete truth. The fact that it treated what a random person said as a definitive fact highlights the unreliability of the whole thing. For all it knows, I could have been making the whole thing up!

 

All the more reason to pay attention to where you're gleaning your information and to only regard reliable, trusted sources on the internet. The phrase, 'you can't make this stuff up' is now subject to suspicion. Yes you can, and that's why this AI stuff is so dangerous. I'm no conspiracy theorist but I will say this. These AI guys are mighty convincing with their rhetoric. Above all......trust but verify.

 

Lol, I get the same thing sometime. I go in and out of certain series, so occasionally I will get back into one and google something, and like the top 3 results are Cointalk posts BY ME! Lol. Guess not too many people are into said fields.

 

Yeah, that is a major problem. All of the hype around AI and the certainty of the language in which its written makes everyone think its authoritative, when its relying on the same old cesspool of internet "information".

What I have always said the last 30 years is simply becoming more so, we have way too much data and way too little real knowledge to interpret the data correctly.