5 Comments

Partially agree. However, in real-world situation, a lot of problem can not be defined precisely and the result can not be evaluated precisely either, like abstract summarization or most of computer vision problems, to apply machine learning in these field, auto-ml just does not work. Other than these, more complicated tasks like reasoning, we are not even close to start working on it right now.

Expand full comment

Not sure if you intended to leave out the topic of input data augmentation,but it has been shown to explain a lot of the recent image classification improvements.

Expand full comment

Small curiosity: do you have any idea of what you'll do next?

Expand full comment

Small curiosity: do you have any idea of what you'll do next?

Expand full comment