If you look at any leaf on any tree branch, it's similar to but not exactly a repetition of the previous branch. So the new science of complexity or showing how an architecture can be produced just as quickly, cheaply and efficiently by using computer production methods to get the slight variation, the self-similarity.

But I do believe architecture, and all art, should be content-driven. It should have something to say beyond the sensational.

The rule seems to be that there are no absolutes, that what is rare is prized. Thus, in times of relative affluence, thin models become dominant.

Pick up a sunflower and count the florets running into its centre, or count the spiral scales of a pine cone or a pineapple, running from its bottom up its sides to the top, and you will find an extraordinary truth: recurring numbers, ratios and proportions.

I was already writing about the idea of a 'multiverse' in the 1970s, though I might have called it the 'pluriverse.' How was I to know it would turn out to be the standard model? Actually, I consider myself an enlightenment fossil.

It's a mark of any icon that it should be open to iconoclasm.

Science is a victim of its own reductive metaphors: 'Big Bang,' 'selfish gene' and so on. Richard Dawkins' selfish gene fitted with the Thatcherite politics of the time. It should actually be the 'altruistic gene,' but he'd never have sold as many books with a title like that.