### Meaningful Math Proofs and “Math is not Calculation”

Sunday, August 5th, 2012Mathematicians are generally thought to be very good at calculation, and they sometimes are, but this is not because math is about calculation, just as astronomy is not about telescopes and computer science is not about computers (paraphrasing Edsger Dijkstra).

But if it is not preeminently about calculation, then what is mathematics about? The common answer, given by mathematicians themselves, is that it is about solving problems–learning to do so, and to think in a logical fashion. I think this may also be misguiding. Let’s take as an example Fermat’s last theorem (which may also be misguiding, as some math is not only about historical unresolved questions with little impact, in some sense, Fermat’s Last Theorem is not much more than a curiosity).

As is well-known, Fermat’s last theorem was solved by Andrew Wiles in 1994. The solution wasn’t simple at all, and it required a lot of convoluted math (Wiles himself got it wrong first and a few years later fixed it). The solution required much more machinery than the machinery required to formulate Fermat’s original statement, it is like going after a fly with a bazooka. For some mathematicians concerned with the foundations of math, Wiles’ theorem may not be as satisfactory they would have wished, even though as a piece of mathematical work connecting several areas of advanced math.

Wiles’ exact solution couldn’t have been arrived at earlier, as it draws upon some of the most sophisticated areas of modern math, it transfers a pure problem of arithmetic to a very powerful framework founded in algebraic geometry and analysis (elliptic curves, modular forms). If Fermat had a solution it wouldn’t have been at all similar to Wiles’. In fact, even before Wiles’ proof, no mathematician ever thought that Fermat’s last theorem was actually false. The consensus was that the theorem was true, so the long expected proof was supposed to shed light on the reason behind this truth rather than simply confirm it. Wiles’ great achievement only confirmed that with a very sophisticated set of tools one could prove Fermat’s last theorem, but it didn’t shed any light on why it was true (if it does, it is still hard to justify such a tradeoff in which to shed some light in arithmetic one requires the use of set theory).

Wiles’ solution doesn’t answer the question of whether Fermat himself had a more basic (but no less valid) proof. The most likely case is that Fermat’s last theorem is actually independent of number theory, and requires the full power of set theory. Hence Fermat very likely had no such solution, contrary to what he claimed.

Fermat’s theorem is, however, a rare case of the ultimate great challenge, and not part of the common practise of the “average” mathematician. The work of an “average” mathematician depends very much on its field. If the field is concerned with foundations it may involve a lot of set theory and logic, which in large part involves finding sets of formulae to prove other sets of formulae and looking for relativisations. In algebra or group theory, or even topology, their work may have to do with all sorts of symmetries or ways to characterise things in different ways in order to establish new and unanticipated connections, shedding more light on one area in terms of another, from a different point of view (and employing a different language). This connection is in fact the greatest achievement of Wiles’ proof of Fermat’s last theorem, as it connects several subdisciplines of modern mathematics building sophisticated intuitions of true statements, ultimately connected to Fermat’s last theorem.

I think math can be better described as a discipline of patterns (e.g. mappings, homeomorphisms, isomorphisms). Patterns are integrally related to the discipline I value most: algorithmic information theory, which is the ultimate field of the study of patterns, patterns in strings and objects, leading to applications such as compression algorithms.

One of the consequences of this change of mindset is that it will allow students access to computers. Computers and humans together can achieve an impressive number of things; they do so everyday in all sorts of fields, from car design to the development of new chips for the next generation of computers. In fact, airplanes today are mostly driven by computers, no pilot or set of pilots alone would be able to fly a modern airplane these days, and more than 60 to 70% (in some markets even 90%) of the operations in any top stock market are computer made, with no human intervention. Today’s industry wouldn’t be conceivable without the combined forces of human creativity and motivation, and the computer’s speed, accuracy and productivity.

There is an important new trend spearheaded by the UK initiative Computer-Based Math

promoting these very ideas. Conrad Wolfram has recently explained in a blog post how computer programming could not only help to teach math, but how they could merge into a single discipline. I may have a chance to talk about all this at the House of Lords soon, where the education shift would need to take official form.