Weren't self-training algorithms with us for the last few decades? I would tend to think that they were just too compute-expensive to be used widely before, and now they are actually getting to work.
You are viewing a single comment's thread from:
Weren't self-training algorithms with us for the last few decades? I would tend to think that they were just too compute-expensive to be used widely before, and now they are actually getting to work.
yes they were, but this goes one step further in that there is much less involvement of programmers. I think the focus here is the amount of data it is needed to be fed for it to become smarter on it's own. It takes a new approach to help AI get smarter on its own.