I believe that AI is taking the wrong path. They are trying to work top down because that’s how humans who are generally successful at solving problems approach solving problems. Problems that are somewhat familiar are best served with a rigorous approach that leads to a planned conclusion. Still, many of the greatest break-throughs have come about by accident, i.e., not following the general path at first. Penicillin and Post-It Notes come immediately to mind.
But humans only solve most problems the usual way because they have a solid ground work to start from. The accidents create a new ground work for development. Accidents come about by not following the normal path, and I think the solution to AI is in taking a different path that will lead to a ground work that will support the entire field.
AI should take the approach of developing siloed expert systems. Make lots of them and keep refining until commoditized. Then start working on higher systems that can merge related systems together though interfaces like web services (but more efficient). Then build ever higher systems until a small set of controlling systems can leverage the legacy systems. The legacy systems, truly failures to create AI and the wrong accepted path will provide an infrastructure that will support a true AI solution.
© Scott S. Nelson