RE: Are we reaching the limits of artificial intelligence?

Andrew J — Wed, 12 Aug 2020 19:03:40 GMT

I think the breakthrough won't come with faster computation but with alternative approaches. At the moment the algorithms are conceived by people with their own mental boundaries. When we are able to create a seed algorithm that genuinely allows a machine to discover its own distinct and separate algorithms, implement and improve on them, and move beyond it's seed, then we'll see some major breakthroughs.

Where the researchers mention 'new method' it seems to me that it isn't a new method, it's just the same method(s) running on faster hardware. I don't see how that enables AI to 'learn complex tasks without supervision' anymore than it could now (but slower.) Whilst AI is bounded by our physical thinking, experience and derivation of algorithms from that I don't see improvement.

I do agree with Dab that this is just beginning to find its place and I suspect that frantic work is going on what I called the 'seed algorithm'. One could have interesting debates on this subject - which suddenly reminds me that I did, back in the 1980s with a friend of mine before we went to University!

RE: Are we reaching the limits of artificial intelligence?

DAB — Wed, 12 Aug 2020 18:12:17 GMT

After over forty years of evolution, AI has a lot more potential use in areas as yet unexplored.

There is a wealth of information hidden within a trained network that has yet to be exploited because the current tools and implementations hide the data and embedded analysis within.

This is not a technology that is done, it is a technology just beginning to find its place in the universe.

DAB