Is the Knowledge in Expert Systems a Good Idea?

In anet line of thinking, disentangled representations, if you might. So for many years, i've thought it's really important that we come up with learning algedems in which the important factors are nicely separated and easy to pick up from the representation. We can maybe just learn with linear models about the things we care about. But this is missing out on a very important ingredient, which classical eye systems can remind us of. There's something really powerful that comes from distributed representations,. The thing that really makes mornets work so well. It's hard to replicate that kind of power in a symbolic world.