Your example of a fruit fly is misinformed. The fruit fly "intelligence" you quote has as much to do with its physical properties as its "brain." The physics of insects is vastly different (weight and lift vs viscosity of air) of any man made flying machines. The evolved fruit fly's brain only has to sense very little to drive a few reactions in order to exhibit complex behavior. But the algorithm for sensing and reacting is not all the complicated or intelligent. If you don't think we can build something to sense a shadow (fly swatter) and jump backwards you are sadly mistaken.

My history is one of developing CAD algorithms. Those algorithms would embody a number of methodologies in order to exceed human design capabilities. Computer "intelligence" has largely been specialized, and been implemented for everything from chess playing, simulations, complex search algorithms for a variety of design issues, speech recognition, visual pattern recognition, medical diagnostics and the whole class of deductive recognition software (eg. fault diagnosis), planning algorithms...

Just as a single example of computers acquiring intelligence, lets look at the chess algorithms. I find it interesting when you break them down, they have 3 components. First, they memorize large numbers of opening scripts. Secondly, the mid game is characterized by evaluating positions, projecting future moves and pruning. Third the end game has some rigorous situational preknowledge and specialized algroithms to deduct and close out the ending. The only difference between humans and the machine would probably be the method of pruning. For the humans, it is largely classifying thousands of the same moves together and lopping them off together. Whereas the computer will follow all moves a certain amount into the future before pruning.

We are just in the earliest stages of integrating many of these specialized "intelligences" into single machines. Those machines have intelligence of maybe insects, but soon it will be mammals and beyond. The truest test of intelligence is the renowned Turing test and can have little or nothing to do with navigating the real world. Does a blind amputee have intelligence?

Today, if you could have a centralized computer system have access to the combined artificial knowledge bases and software in existence, it would have formidable abilities. You could not beat it in chess, and there would be few engineer design problems that it could not simulate, and solve. It would have encyclopedic knowledge of everything, and be able to quote baseball stats at its leisure.

The concept of the book is probably just that as computational density approaches that of the human brain, then you can have artificial intelligences correspondent with human abilities. That same thought occurred to me about 25 years ago. I don't buy the quantum computing argument of the book, as that is in the far future for massive number of states. Right now I think we have built quantum processors having about 10 states before it disassociated.

Anyway, single machine "intelligence" is definitely "evolving" and a great example are Japanese "toy" robots. They are developing robots that simulate pets and eventually they will be simulating human companions. I believe you can simulate any human behavior, including emotions (situational and predictable reactions). If a computer tells you it is scared, who are you to argue? Can you argue with a human who tells you they "feel" something? That is their own experience and we can only empathize.

I believe we already embody higher intelligence in that we are part of a social "organism." This is the so call mass mind effect and is why markets exhibit more intelligent behvior than individual human transactions. This mass mind is evolving (specializations, improved communication, knowledge base increasing, etc.) and is not unlike the evolution form single celled organisms to multicelled (e.g. the first ganglia, the spinal cord...).

Anyway, this can become quite a long debate, so I will prune my thoughts.