I think organisms have to be computed from their DNA sequences, just as we have to compute proteins from its DNA... a protein from its DNA sequence. Now, when I say compute, I... I mean it in the sense of... of a computation, this need not be a calculation but in fact... and we may not even use a... a digital computer, almost certainly we will, because it's a universal computer, when programmed with the right program we could do this. Now, we have to do it in this way for the simple reason that this is the only way you can get to the final... the so-called final explanation. Otherwise we simply have a set of anecdotes, and this is particularly true if we want to go to yet a more complex thing, which is how to account for behaviour, for the operations of the nervous system. If we say what does an animal or a person do in such and such circumstances, we simply make observations of those circumstances, we then try to look at relationships between these and then of course we try to get a predictive theory of behaviour, and in many cases it falls... it... it fails when it becomes a little bit richer because this simple connection is usually too simple...too simple a correlation. Now, rather than providing an anecdote, roughly speaking, to take each case and have a look; that is if you ask what does this animal do if you give it a cigarette to smoke? Well, you have to give it a cigarette to smoke and see what happens, then you can give an answer to that question. But there is no end to asking such questions, so we have what I call a halting problem in the Turing sense. There is no end to the number of questions and answers that we may pose about any complex system. Now, the only way then we could answer those questions without recourse to the system itself  — which of course scientists would say is description — is of course to have a computable theory of it, so that if you can ask me that question I will get the theory to produce the answer. We could then go and check that it's correct. Now, if you have a unique theory that can do this, and I actually don't think that for behaviour this will be attainable in a very general sense, because — that is, from the genes — because I think what we can have is a unique theory about how the nervous system is constructed from the genes, then we have to start from... with nervous systems and then develop theories of how they work. That is why with the C. elegans [Caenorhabditis elegans]project it became important to specify the wiring diagram and to ask the question is behaviour computable from the wiring diagram alone — plus some initial conditions, I should add — because that's a style of answer which is very similar to answering questions about how you build machines, is it computable? And this I think raises the question of what is considered a proper computation in this sense, and that I think that can be answered very quickly. That is, it should be in the machine language of the thing being simulated. That is, if you wish to describe nematode behaviour it should not be in terms of sin(theta), cos(theta), which may well be very good to generate the wavy... the serpentine motion that they have, but must be in terms of neurones and their connections. That is the machine language of that. Just like the machine language of development is in terms of cells and the recognition proteins they carry on them, and all the mechanisms they have to process signals. That's the machine language of development. Machine language of development is not gradients and it's not differential equations. So that is... so I just ruled them out instantly as just boring descriptions of what there is in an incomprehensible language, just like sin(theta), cos(theta) is another boring description in an incomprehensible language. So the machine language of the object is important and it is, if you like, the task of experimental biologists to define that.