Estimating the intelligence of systems sees us resorting to crude computations of neural count, processing speed and connectivity. So we as a species have around 10 neurons and a 10 connectivity. Given the uncertainties in getting the numbers right in the first place, counting the numbers per unit volume of brain puts us in the 10 bracket.

Now include our synaptic processing rate of around 10 to 100 bits per second and some localisation assumptions, and we look like a 1 teraflop per second processor. But we know we are much more.

Chances are that our current estimates of our brain capacity are a grossly on the low side, if for no other reason than we do not fully understand how our own wetware functions. One obvious factor that does distinguish us from silicon machines, is our relative information in/out rate. For us, information flow is dominated by input. From the moment we are born until we die our sensory system is feeding our brain vast amounts of information. Our utterances and animation pale into insignificance in raw bit-rate terms compared with the visual, audio, and tactile input we constantly absorb.

As best we can estimate each of our eyes feeds us bits from 127 million rods and cones in the retina, via the visual cortex, at about one gigabit a second. This far outweighs the feed from our hearing, tactile, smell and taste sensors which total less than 20 megabits a second. In contrast, our typing rate is only around 40bit/s, speech is about 100bit/s, while gesticulation and facial expression are even lower, but of course can be socially encoded to covey far more. A glance from a loved one at a cocktail party, or a close colleague in a meeting, can be worth megabits.

Contrast all of this with the average PC, or better still, the mainframe. We complain about them being dumb machines, and yet we deprive them of input. For the most part all they get is a diet of low bit-rate alphanumerics. No direct visual and audio feed, and certainly no tactile taste or smell. From billing customers, to issuing pay slips, predicting the weather, or controlling a space shot, their input is meagre compared with their output. No wonder they are dumb.

For machines to achieve true intelligence might then require a change of I/O bias. I suspect they will need much more input data, more contemplation processing time, and less time and resources devoted to output. Often thinkers are accused on being non-communicative, but perhaps that is the secret. Deep thought demands concentration and a degree of mental isolation. Input dominates in all such cases, but when the output comes it can be devastating.

We have within our grasp the basic technologies to build machines that might just be capable of some form of free thinking. What is lacking is a suitably formatted information feed, a sensory system that will match them into our, and other worlds. Perhaps that is why one famous machine of science fiction came to the conclusion that the answer to an important human question was 42, which is binary 101010 - just a free running clock.

Peter Cochrane holds the Collier Chair for the Public Understanding of Science & Technology at the University of Bristol. His home page is:
http://cochrane.org.uk