I put ChatGPT's new vision feature to the test with 7 prompts — the result is mindblowing

[admin]

OpenAI's GPT-4o is impressive and with it comes a upgrade to ChatGPT's vision capabilities.

I put ChatGPT's new vision feature to the test with 7 prompts — the result is mindblowing : Read more

 

[husky91]

I guess I'd be interested to know how it does with real images rather than AI generated ones. Can it identify not just what the image is but specifics about the image. Like with a picture of a building, can it identify the building and facts about it? Do this with exif data stripped out and see if it can identify the event like a picture of the stadium during Super Bowl 50 or whatever. Can it identify the opponents and date?

 

[Red_Tom]

Do this with a real image and getting the same result and I'd be impressed. How do we know that GPT-4o in this article's scenario isn't just reading meta-data/prompt text from the Ideogram AI images, and not actually recognising things in the images? You mentioned using it at an event in Paris but elaborated no further on this - how did you use it in Paris and what were the results there?