Gelernter's thesis is that the goals of the Artificial Intelligence programme cannot work, because AI misses out a crucial form of human thought, different in kind from our symbolic, analytical thought. He explains that our thoughts fall along a 'focus' spectrum, from high focus analytical thought, through low focus, analogical thought, and even further, to dreaming. And it is only when we engage in low focus thought that we can make creative leaps.
So far, so good. But he then says that thought at the low focus end of the spectrum is necessarily emotional, and that since emotions require a body to experience them, that AI cannot work without embodiment. Here I disagree with his first step: from introspection, my own low focus thought (which I now no longer have to call 'woolgathering') does not seem to be emotional. Whenever I have recreated a stream of low focus thought, I have always found linking events that make sense without having to invoke an emotional explanation. In particular, whenever something "breaks my dream", it is always a common event, never a common emotion, that does so.
I am sympathetic to the idea that intelligence requires embodiment. It may be the case that in order to develop certain kinds of intelligence we need to be situated in and interact with our environment in complex ways involving a lot of feedback loops. [For example, some hearing children of deaf parents had access to a TV, but never interpreted the sounds as speech. This could be due to the lack of feedback: lots of things make noises; maybe it's because those noises change in response to noises we make that we realise they are interesting and important noises.] But I remain to be convinced that any need for embodiment is due to a need for an emotional response.
However, Gelernter's idea of a spectrum of thought is interesting. Especially the idea that creative, low focus thought can be encouraged by engaging in an activity that precludes high focus thought: all of one's concentration is required for high focus thought, so if some concentration is engaged in performing some routine task, such as driving, then low focus thought is the only remaining possibility. This may explain why I sometimes have good ideas while playing 'mindless' computer games!
This is an impassioned plea to take aesthetics seriously in computing, both software and hardware. Beauty in technology lies in a fusion of simplicity and power. As Gelernter points out, most top rank mathematicians, scientists and engineers emphasise the importance of beauty as a guiding principle for their subject. So why isn't this claim taken seriously? He has a suggestion, using the example of the Mac desktop versus Windows: elegance isn't manly. Real computers are built by men, for men. He explores this theme in a series of interesting case studies, including Turing machines, Algol, Unix, Xerox PARC, and Java, and some chokingly funny acerbic little examples.
At one point Gelernter says we don't need more mathematics in software development, we need more aesthetics. Although on the one hand, I know what he means (that the problem doesn't lie in lack of rigor, but in lack of elegance), on the other, this statement looks a bit bizarre: after all, mathematics is almost all aesthetics: if it's not beautiful, you aren't doing it right.
Separating a heartfelt rant about ugly software from an equally heartfelt rant about ugly hardware, the middle of the book contains an interesting digression on Lifestream. This is a simple powerful concept Gelernter and his co-workers have invented as the next step, breaking through beyond the desktop metaphor, as a new way to organise computer files, emails, electronic calenders, TV, and phone calls. This breaks away from the old ways of doing things -- we use hierarchical file systems to help us find our files. Now that computers are so powerful, why not let the machine find the file -- by meaningful content or context, not by some artificial filename, or by some rigidly imposed and unrealistic single hierarical classification scheme. [Peter Cochrane has a similar insight about email folders in his Tips for Time Travellers.]
So, there is much food for thought in here. What would an aesthetic-aware computing industry look like? We should try it. It certainly couldn't make things any worse, and it might improve our lives beyond all recognition.