A Way Forward
- Given that symbolic AI has a flaw at its base (unground symbols), and
- subsymbolic AI solves only isolated problems,
- how can we move forward?
- The obvious thing to do is combine the two.
- Of course there has been a lot of work in doing this, but
it has not been very successful.
- However, we have an example of a system that combines the
two quite successfully. Humans! (and other biological systems)
- Newell thought we were symbol systems, and we clearly are
neural systems.
- To move forward, lets build systems that are based on neurons,
human brain topology, and can match psychological tests,
- and get them to do smart things.
