Inventor of quantum computing, and advocate of the many-worlds interpretation, David Deutsch is interviewed in this week's issue of New Scientist (9th December 2006, p51). It's a short interview, but Deutsch makes the following crucial point: "In fundamental physics research, progress is only made when people address problems within existing theoretical physics...In the last few decades, many theoretical physicists have assumed that further progress can only come from looking at new mathematical models and then wondering if the models are true representations of nature. An example is string theory.
"I think it's unlikely that a research programme of that kind can work. Even if you found the right mathematical object, you probably wouldn't even recognise it because you wouldn't know how it corresponds with the world...I would warn against expecting the answer to come from a new mathematical model. It should be the other way round: first find what you think might be the solution to a problem, then express it as a mathematical model, then test it."
Deutsch is spot-on. In very simple terms, a theory of mathematical physics can be broken into (i) an empirical domain, (ii) a collection of mathematical structures, and (iii) a set of correspondence rules which link parts of the mathematical structures with parts of the empirical domain. String theory is an attempt to invent a new fundamental theory by exploring only one of these three dimensions: namely, the collection of mathematical structures. Hence, despite string theory's notorious proposal that space-time has many more dimensions than we can currently detect, string theory itself is rather one-dimensional.