The Folly of Understanding: When AI Outsmarts Us | 05 мая 2022 года, 00:59

I always thought I can explain how anything in the world is built, at least to some extent. I always thought that I knew how machine learning works. I was even creating a chess recognition app which worked well. But the latest news from that area are really shocking. I can’t explain how the trained models can produce such wonderful results.

Look, this model explains the jokes and untwines the stories. I feel like I am being fooled. It’s too good to be true.

https://arxiv.org/pdf/2204.02311v2.pdf

P.s. DALL-E 2 is already great https://openai.com/dall-e-2/ it creates realistic images and art from a description in natural language.

Оставить комментарий