As computer hardware and software have coevolved, whenever a new feature is required its implementation in both hardware and software always seems to bring the best results. Take virtualization for example, initial virtualization mechanisms were implemented in pure software models. It was not until the advent of this technology in hardware (e.g. Intel VT) that virtualization became more reliable and more efficient. Of course, software was also changed in order to accommodate this new hardware. In the current state of affairs virtualization is expanding and maturing as both hardware and software coevolve to meet the demands of its human overlords.
“Machine Thought,” I believe will come in the same way, the first glimpse of intelligence will come from a software model. This model will give us insight on how complex systems work and will spawn a number of inventions in hardware that will alleviate many of the flaws that the software model will suffer from. It is through the coevolution of both hardware and software that new intelligent systems will begin to be born. This may require that the line between software and hardware be blurred even further. It may be that the current materials with which hardware is built are not sufficient for expressing the systems that we want to build. New materials may come from biological and/or chemical constructs that may eventually allow other types of computation, maybe even lead to “thought” outside the human mind.