Sunday, September 10, 2017

[SHARE] - Có phải Deep Learning giải quyết hết các bài toán AI?

Trả lời của Prof. Yoshua Bengio - tiền bối về deep learning cho câu hỏi: Why aren't artificial neural networks used for everything? Of course you wouldn't do colliding galaxy simulations with them or anything ridiculous; but it seems that in AI, they have a very good applicability range.

If you have enough prior knowledge on a problem you don't need any learning, or it is enough to design the appropriate fully observed graphical model (you know what all the relevant random variables are and you either know their relationship or observe them in enough data), or just learn the "top layer"  (e.g. with a good feature space or a good similarity/kernel you can just plug your favorite SVM brand).

Another basic reason why I believe that artificial neural nets are not used for "everything AI" is simply that we are far from having captured from the cortex and from our ingenuity a learning algorithm as powerful as biological brains (in particular primates', and to some extent mammals and birds) enjoy, i.e., more real fundamental research is needed!