DISCUSSION 
words society should be told that his prediction, however well- 
calculated, cannot state an objective “‘certainty”’, but only 
something whose validity depends on their attitude to it. His 
role is quite different from our normal function as scientists, 
where we seek to make objective predictions which stand 
independent of the attitude the recipient takes to them. 
A second point concerns the limitation of machines. As I said 
before, once the kind of performance required can be completely 
specified, it is in theory possible to construct a machine to carry 
it out. Such machines may be inordinately complex, but they 
can be surprisingly “‘original’’; already for example, machines 
have been programmed which were able to produce novel 
proofs of simple theorems. The difficulty arises when we ask 
ourselves how to specify a value judgment as made by a human 
being, as distinct from a mechanical deduction from data. 
How do we decide what ought to be done, as distinct from 
what can be done? At this stage we do not begin to know the 
answer. If we did, and could make explicit the behaviour 
pattern that would make a machine an acceptable arbiter, that 
limitation would disappear. 
The ultimate question is this: granted that our brains are 
machines, is it possible for any machine to discover and embody 
within itself a full description of itself? I think it was von 
Neumann who first raised this as suggesting a likely limit to our 
power to synthesize human behaviour. It is not that we cannot 
get a machine to do anything that has been specified, but that 
as men ourselves, our understanding of what it is to be a man is 
likely to be fundamentally limited and therefore a complete 
specification will always clude us. 
Lipmann: I have little doubt that machines will eventually 
be able to do everything. The problem may really be a 
psychological one, namely, that we do not want to be left 
without mystery. An unsolved complexity has a mysterious 
quality and after resolving it we are dismayed to find another 
mystery has been ‘‘lost”’. I think that in the working stage it 
might be more helpful not to assume that there will always 
remain a residue that we cannot understand. 
180 
