³ÉÈËVRÊÓƵ

David Salt falls into the same trap as Roger Penrose when attempting to understand connectionist software, when he suggests that the algorithms are a limitation to computer systems simulating the brain (THES, April 28).

The biochemical "algorithms" that dictate the internal behaviour of the human mind are the analogue of these limitations. Neural networks and the no doubt, more complex models that follow them, will better and better model these "biochemical algorithms". The art of computing is modelling and simulating the underlying behaviour of any system. What makes the new network algorithms special, is that they can change with experience.

Whether all processes of the human mind can be modelled, is debatable but I would hypothesise that consciousness and intelligence have some underlying behaviour that can be modelled at certain levels. No one can state they will produce, or understand fully, human intelligence or human consciousness but there will be similar models which may give insights into understanding these concepts. Only bigots have the audacity to suggest that we shall never achieve any scientific goal, they do not see that no one can predict what developments the future may bring.

Though the philosophical debate is interesting, the real work is the development of evaluation of better and better models. That evaluation must include comparisons with human consciousness and intelligence. The philosophers would be better trying to identify consciousness or intelligence before they can decide whether they are achievable. The question for those developing intelligent systems is not whether we will achieve but what we can achieve.

MARTIN LEFLEY

Department of applied computing and electronics

Bournemouth University