I came across a very interesting debate on mathematical proof. The debate started with a small subgroup of mathematics who are strong believers in computer verification of proofs. This quest is promoted as the QED project (check out the QED-manifest). In order to verify a proof by computer, it needs to be recoded, because the human proof takes simple steps for granted. The computer proof is a ‘computer program’ whereas the ‘human proof’ is more a formalised story.
The debate is interesting for many reasons and I think it has a broader meaning than maths only. In a sense it is about what we still think is human. We embrace the ‘romantic side’ of maths: the beauty in a proof. I am not a mathematician, but I still remember the beauty of a good proof as I learned in high school. There is beauty in the idea of a solid, water-tight proof but there is also beauty in the steps. The set of different rules, the one leading to another is an art. My son is studying to be a maths teacher and he told me about the wonderful feeling he has if he succeeds in proving something, even if this has been proven already by a great number of students. Needless to say that the feeling of proving something new must be euphoric. Now you could interpret this QED-project as an attack on the romantic side. If there is any feeling and emotion attached in constructing or understanding a proof, the computer version of it is devoid of that. A simple proof that even I can understand is elegant in ‘language’ but not any more in computer language (check out this pdf to see the difference. It is in Dutch but I guess that anyone with algebra background can read it no matter the skills in our Low Lands Language).
In the Dutch daily NRC Bennie Mols wrote an article about the subject. What really struck me in this discussion is the standpoint of Hendrik Lenstra (a Dutch mathematician). He claims that the act of proofing and judging a proof is basically a human act, based on pattern recognition. The act making and checking a proof helps you to better understand the subject. It evokes ‘knowledge’ as an internal phenomenon. He shares an interesting and wonderful story about a theorem he was trying to understand. He wanted to know if it was correct. He had it both computer-checked and asked another mathematician to check it. He consulted an expert in computer mathematician. The computer had to work a day on it, and showed the proof was right. The human expert got about it in a different way. He looked at the theorem and said ‘oh I can calculate that for you’. Within in few minutes he had the maths done – writing it out on a piece of paper. His conclusion: the theorem is wrong. Of course – sorry for those with a romantic view – the expert had made a small miscalculation. The computer was right. The interesting thing though, was that actually mr. Lenstra learned a lot more from the human interaction. Checking the piece of paper, understanding the theory used by the expert gave him deeper understanding about the problem. Sharing this with the computer expert, enabled him to rewrite the computer proof in a much more elegant way, so the computer now needed a few minutes as opposed to the former day of computer time.
I see many learnings is this beautiful story. Too many to ponder about in my blog that I want to be short (actually this posting exceeds my intended maximum manifold, I hope to reach a sort of Haiku blog in the future, revealing basic truths within 11 words. But I haven’t yet reached that skill). Let me highlight some:
1
whether we like it or not, automation of knowledge is bound to grow. Even the critics of QED see that it can be extremely useful to check extremely complicated proofs that human race is not able to verify with 100% certainty
2
This will not squeeze the romance out of our lives. As long as we are here on the planet (could be shorter than intended, but let’s assume we will survive the next 2.000 years) the field of ‘beauty’ will change. Beauty is in the eye of the beholder.
3
Knowledge and understanding is the field of humans. Up to now the computers can assist.
3
In the end it is all about the human factor. The QED is a sort of ‘bookkeeping’. Converting proofs into computer language is not like hot science, it is more the work of a monk. But that can be very satisfying. I can imagine the writer of the pdf I referred to above in his feeling that programming a proof is a wonderful thing to do.
4
the frontier on automation is ever changing. The discussion is now about the verification of proofs. Could it ever shift towards production of proofs? Maybe then computers would become like humans more. So that would not really shift things. We would still have the two different states; the understanding state (which is emotional by definition) and the calculating mode. The shift would mean that the emotional state (up to now the privelige of human and probably animals) is now available to computers as well. We are not near that in the least.




