I have this feeling about edge.org that they manage to invite remarkable people to have incredibly shallow content that is full of buzzword (maybe this is what happens if you try to pull off profound things for public). Just an experience.
"The realization from that is that the thing I've spent a large part of my life doing, which is building computer languages, is not such a bad idea. In a computer language, you do get to represent more sophisticated concepts in a clean way, which can be progressively built up in a way that isn't possible in natural language.
One of the things that I'm interested in is how we communicate goals to AIs. How do we talk to the AIs? My basic conclusion is that it's a mixture. Human natural language is good up to a point, and has evolved to describe what we typically encounter in the world. Things that exist from nature, things that we've chosen to build in the world—these are things which human natural language has evolved to describe. But there's a lot that exists out there in the world for which human natural language doesn't have descriptions yet. Even though our AI systems might effectively find those descriptions, we don't have ways to say those ourselves."
I really try to understand Wolfram's point of view but always end up completely disagreeing with him. He is so blatantly biased to wanting everyone to use his baby Mathematica that his reasoning doesn't make any sense. My arguments:
1. For starters, ZERO programming languages can represent a concept in a more sophisticated way than natural language, simply because we don't encode understanding in code: where a computer sees a loop condition, followed by an if and a return statement, we see the real meaning 'check if this item is in the list'. The meaning is so much more useful because it tells us the WHAT not the HOW. By know the what, we can understand what is trying to be accomplished, and so be able to predict behaviour, see logic errors the compiler can't see, make modifications to add other features, change the entire algorithm if the data structures change (use hashtable perhaps) and so on. The fact that Wolfram doesn't grasp this point, makes me seriously consider he doesn't have a clue what he's talking about.
2. The next most important thing about natural language is it is not sequential, like code. There is no structure for one precise all-important reason; you do not know what you are talking about! How many times have you had a conversation with someone, for them to then tell you that you are wrong, or perhaps add modifications to what you just said? Natural language facilitates a way for multiple nodes (humans) on a network (social groups) to share information that is not fully correct, and is expressed in such a way that another node on the network can pick it up, and express modifications, resulting in a final timeline that the group can agree on. Algorithms written in code on the other hand are BRITTLE, and expected to be correct every step of the way. The fact that no programming language can agree on a single way to handle exceptions/errors in production systems is simply a byproduct of them being unable to adapt like natural languages can. It really shows off the incredible power of our natural languages, and I hope I'm alive to see the day we crack natural language understanding!
