Testing AI image descriptions

[Maia]

Post a picture for me and I'll run it through ChatGPT and Grok image analysis, and then you can tell me if they're any good, and which is better.

Please nothing rude, though, as I don't want to be blocked, or anything.

AI is suddenly everywhere, so I thought I might as well do something useful with it.

[accelafine]

You can't post pictures on here any more.

[Maia]

You can't post pictures on here any more.

That's a pity. I didn't know that.

[Flannel Jesus]

You can post links to pictures though. I'll find one for you maia.

https://dspncdn.com/a1/media/692x/a7/37 ... 8d25cf.jpg

[Maia]

You can post links to pictures though. I'll find one for you maia.

https://dspncdn.com/a1/media/692x/a7/37 ... 8d25cf.jpg

ChatGPT
+++This image appears to be a digitally manipulated or distorted black-and-white photograph of a person in a military-style uniform. The face has been altered to create a surreal and unsettling effect, with the mouth exaggeratedly stretched and the hair seemingly elongated or blurred in motion. The result is a grotesque and uncanny appearance, possibly intended for artistic or satirical purposes. Let me know if you need further analysis or interpretation.+++

Grok
+++This is a black-and-white photograph showing a side profile of a person with short, neatly combed hair. The individual is wearing a collared shirt or jacket, and the image appears to capture them in a moment of intense emotion, possibly speaking or shouting, as their mouth is open wide and their facial expression seems animated. The background is dark, which contrasts with the lighter tones of the person's face and clothing, drawing attention to their profile. The overall quality of the photo suggests it might be historical or archival in nature.+++

[Flannel Jesus]

You can post links to pictures though. I'll find one for you maia.

https://dspncdn.com/a1/media/692x/a7/37 ... 8d25cf.jpg

ChatGPT
+++This image appears to be a digitally manipulated or distorted black-and-white photograph of a person in a military-style uniform. The face has been altered to create a surreal and unsettling effect, with the mouth exaggeratedly stretched and the hair seemingly elongated or blurred in motion. The result is a grotesque and uncanny appearance, possibly intended for artistic or satirical purposes. Let me know if you need further analysis or interpretation.+++

Grok
+++This is a black-and-white photograph showing a side profile of a person with short, neatly combed hair. The individual is wearing a collared shirt or jacket, and the image appears to capture them in a moment of intense emotion, possibly speaking or shouting, as their mouth is open wide and their facial expression seems animated. The background is dark, which contrasts with the lighter tones of the person's face and clothing, drawing attention to their profile. The overall quality of the photo suggests it might be historical or archival in nature.+++

ChatGPT understood some interesting stuff from the image that Grok missed.

It's actually a picture of who I believe to be Mao Zedong, it's an old and grainy picture, and it's been distorted to be just a little bit terrifying. His teeth are elongated, things are stretched, other things and squished, he looks alien. He's making what appears to be a happy face, but the distortions turn it into a terrifying joy. The smile of an evil man (which I guess many would argue is what he was anyway, without these visual distortions).

I actually didn't realize it was Mao Zedong at first, I knew it was a famous asian political figure. I asked ChatGPT myself about the image, and its first answer to me - like yours - gave no identifying information. I asked ChatGPT, "Who do you think this is? I think it's Kim Jong Un" and it said "I believe it's more reminiscent of Mao Zedong". As soon as I read that, I realized, ahh yeah that's the name that wasn't coming to me. So it's interesting that ChatGPT was able to figure that out from an extremely distorted image.

[Maia]

You can post links to pictures though. I'll find one for you maia.

https://dspncdn.com/a1/media/692x/a7/37 ... 8d25cf.jpg

ChatGPT
+++This image appears to be a digitally manipulated or distorted black-and-white photograph of a person in a military-style uniform. The face has been altered to create a surreal and unsettling effect, with the mouth exaggeratedly stretched and the hair seemingly elongated or blurred in motion. The result is a grotesque and uncanny appearance, possibly intended for artistic or satirical purposes. Let me know if you need further analysis or interpretation.+++

Grok
+++This is a black-and-white photograph showing a side profile of a person with short, neatly combed hair. The individual is wearing a collared shirt or jacket, and the image appears to capture them in a moment of intense emotion, possibly speaking or shouting, as their mouth is open wide and their facial expression seems animated. The background is dark, which contrasts with the lighter tones of the person's face and clothing, drawing attention to their profile. The overall quality of the photo suggests it might be historical or archival in nature.+++

ChatGPT understood some interesting stuff from the image that Grok missed.

It's actually a picture of who I believe to be Mao Zedong, it's an old and grainy picture, and it's been distorted to be just a little bit terrifying. His teeth are elongated, things are stretched, other things and squished, he looks alien. He's making what appears to be a happy face, but the distortions turn it into a terrifying joy. The smile of an evil man (which I guess many would argue is what he was anyway, without these visual distortions).

I actually didn't realize it was Mao Zedong at first, I knew it was a famous asian political figure. I asked ChatGPT myself about the image, and its first answer to me - like yours - gave no identifying information. I asked ChatGPT, "Who do you think this is? I think it's Kim Jong Un" and it said "I believe it's more reminiscent of Mao Zedong". As soon as I read that, I realized, ahh yeah that's the name that wasn't coming to me. So it's interesting that ChatGPT was able to figure that out from an extremely distorted image.

Interesting. So ChatGPT is better, at least in this instance?

[Flannel Jesus]

Yeah it definitely described the emotional vibe of the image better. Grok just said "intense emotion" and didn't seem too realise it was edited, and didn't pick up on the horror of the image.

[Maia]

Yeah it definitely described the emotional vibe of the image better. Grok just said "intense emotion" and didn't seem too realise it was edited, and didn't pick up on the horror of the image.

I'll bear that in mind for future reference. I'm still quite reluctant to trust them entirely, though.

[Flannel Jesus]

Yeah it definitely described the emotional vibe of the image better. Grok just said "intense emotion" and didn't seem too realise it was edited, and didn't pick up on the horror of the image.

I'll bear that in mind for future reference. I'm still quite reluctant to trust them entirely, though.

Yeah it'll be a long time before they describe it as a human would, but they're remarkably close already, especially gpt. Try deepseek as well, see what it says. And Claude if you can.

[Walker]

You can post links to pictures though. I'll find one for you maia.

https://dspncdn.com/a1/media/692x/a7/37 ... 8d25cf.jpg

ChatGPT
+++This image appears to be a digitally manipulated or distorted black-and-white photograph of a person in a military-style uniform. The face has been altered to create a surreal and unsettling effect, with the mouth exaggeratedly stretched and the hair seemingly elongated or blurred in motion. The result is a grotesque and uncanny appearance, possibly intended for artistic or satirical purposes. Let me know if you need further analysis or interpretation.+++

For comparison, could you ask this AI to describe the image again, using e-prime?

[Walker]

You can post links to pictures though. I'll find one for you maia.

https://dspncdn.com/a1/media/692x/a7/37 ... 8d25cf.jpg

ChatGPT
+++This image appears to be a digitally manipulated or distorted black-and-white photograph of a person in a military-style uniform. The face has been altered to create a surreal and unsettling effect, with the mouth exaggeratedly stretched and the hair seemingly elongated or blurred in motion. The result is a grotesque and uncanny appearance, possibly intended for artistic or satirical purposes. Let me know if you need further analysis or interpretation.+++

For comparison, could you ask this AI to describe the image again, but this time writing in e-prime?

e-prime = compulsory scientific thinking

[Maia]

Yeah it definitely described the emotional vibe of the image better. Grok just said "intense emotion" and didn't seem too realise it was edited, and didn't pick up on the horror of the image.

I'll bear that in mind for future reference. I'm still quite reluctant to trust them entirely, though.

Yeah it'll be a long time before they describe it as a human would, but they're remarkably close already, especially gpt. Try deepseek as well, see what it says. And Claude if you can.

Thanks, I'll explore further.

[Maia]

You can post links to pictures though. I'll find one for you maia.

https://dspncdn.com/a1/media/692x/a7/37 ... 8d25cf.jpg

ChatGPT
+++This image appears to be a digitally manipulated or distorted black-and-white photograph of a person in a military-style uniform. The face has been altered to create a surreal and unsettling effect, with the mouth exaggeratedly stretched and the hair seemingly elongated or blurred in motion. The result is a grotesque and uncanny appearance, possibly intended for artistic or satirical purposes. Let me know if you need further analysis or interpretation.+++

For comparison, could you ask this AI to describe the image again, using e-prime?

I didn't know what that was, but this is what ChatGPT came out with.

+++This image appears to show a black-and-white photograph of a person in a military-style uniform. Digital manipulation seems to have altered the face, making the mouth look unnaturally large with an exaggerated grin. The hair seems to blur or stretch backward, giving an impression of motion or distortion. The overall effect creates a surreal, unsettling atmosphere, possibly suggesting artistic or satirical intent.+++

[Flannel Jesus]

For comparison, could you ask this AI to describe the image again, using e-prime?

What does that mean? Using e-prime? I googled "e prime ai", Google only thinks it's an electric scooter.

Edit. Chat gpt knows what it means. It's apparently just a stipulation that you word things while avoiding all forms of the verb "to be".