ChatGPT

WBahn

Joined Mar 31, 2012
32,854
What the fruck. That's interesting. I see they did get "Colorada" correct though (ha ha ha). Looks like Washington is written in greek :)
I can guess that it was given a picture with low resolution to work from so it could not make out the lettering properly.
I also saw lettering like that somewhere else too and was wondering where it might have come from. Now I know.
It's not so much an issue of resolution on the lettering as it is a much more fundamental issue of image generation versus text generation.
 

MrAl

Joined Jun 17, 2014
13,707
It's not so much an issue of resolution on the lettering as it is a much more fundamental issue of image generation versus text generation.
Hi,

Oh I was referring to the image that might have been given to the 'ai' bot. If it is of low resolution, it will read in very strange characters. It's funny actually, aka "Clown Bot".

Gives me the thought of how many 'ai' bots can you fit into a virtual Volkswagen :)
 

Attachments

WBahn

Joined Mar 31, 2012
32,854
Hi,

Oh I was referring to the image that might have been given to the 'ai' bot. If it is of low resolution, it will read in very strange characters. It's funny actually, aka "Clown Bot".

Gives me the thought of how many 'ai' bots can you fit into a virtual Volkswagen :)
But the AI tools aren't reading characters when they are fed images. Similarly, even when they are generating images with well established text content, they mess up the image representation of the characters because of how they work at a fundamental level using diffusion models.
 

MrAl

Joined Jun 17, 2014
13,707
But the AI tools aren't reading characters when they are fed images. Similarly, even when they are generating images with well established text content, they mess up the image representation of the characters because of how they work at a fundamental level using diffusion models.
Oh that's interesting I never had a problem with that, just that they take too darn long to read an image, so I kind of stopped trying that.
I did have a couple that it could not read though.
 

nsaspook

Joined Aug 27, 2009
16,328
Yes, you get bigger firecrackers.

"It's like thinking we can make an explosion like a nuclear bomb can, by improving gunpowder technology"

https://edition.cnn.com/2025/08/14/business/chatgpt-rollout-problems

The latest ChatGPT is supposed to be ‘PhD level’ smart. It can’t even label a map

For critics like Marcus, the GPT-5 flop was a kind of vindication. As he noted in a blog post, other models like Elon Musk’s Grok aren’t faring much better, and the backlash from even AI proponents feels like a turning point.

When people talk about AI, they’re talking about one of two things: the AI we have now — chatbots with limited, defined utility — and the AI that companies like Altman’s claim they can build — machines that can outsmart humans and tell us how to cure cancer, fix global warming, drive our cars and grow our crops, all while entertaining and delighting us along the way.

But the gap between the promise and the reality of AI only seems to widen with every new model.
 

BobTPH

Joined Jun 5, 2013
11,518
Self-driving cars is 90% a simple physics problem we solved back in the vacuum tube electronics days, it's the last 10% that requires interfacing with irrational humans that's hard.
My point exactly. They can’t seem to get beyond that 90%. If a human could only drive safely 90% of the time they would lose their license, assuming they were still alive.

I think the current AI LLM models are nearly as good as they are going to get, and the next leap will require an entirely different approach.
 

Futurist

Joined Apr 8, 2025
759
I find smut, sexual innuendo inappropriate on a technology and educational forum. There are countless options for participating in such banter, never forget who might read our posts, Young impressionable minds should be respected, childhood should be respected. You wanna post smut, head over to Twitter.
 
Top