|
|
|
|
|
|
by elpocko
385 days ago
|
|
|
I think they've given up on this, it's ubiquitous. I've recently noticed that a CLIP model (a neural network for image understanding) gave me image descriptions using "it's" where "its" would have been correct. Hundreds of thousands of dollars to train the thing; no one noticed, no one cared ¯\_(ツ)_/¯ |
|
|