[LOCKSUIT]

https://openai.com/blog/image-gpt/

:)))))))

[LOCKSUIT]

This is the simplest notebook I've found. Easy to try it yourself. Just change the image to your image link by posting your image to https://imgbb.com/ and then right click it into a new tab then you get the .jpg at end link - a real link, then swap it in for yours in the code.

My first try is below. I'm not sure why the image seems a bit distorted. But it's true so far to its word.

edit: oh, it seems it is 32x32, not the 64x64 they used. My analysis shows this is true.

https://www.youtube.com/watch?v=nlp9ouTSZ4Y

[LOCKSUIT]

Oh....
https://colab.research.google.com/github/apeguero1/image-gpt/blob/master/Image_GPT_Sample_with_Conditioning.ipynb

[LOCKSUIT]

Another impressive AI:
https://www.youtube.com/watch?v=cgakyOI9r8M&feature=youtu.be

my tests:
https://ibb.co/802YPnc

[LOCKSUIT]

https://www.reddit.com/r/OpenAI/comments/hs9obh/which_is_more_interesting/

Interesting to see what the minds of people are thinking on average.

That big post is mine. I felt GPT-2 was more impressive because it captures more human-thought than single images do. And more "tasks".