Lynch is founder and CEO of Autonomy, a maker of software that ties together all kinds of unstructured information -- what's now known as knowledge management. The term wasn't in heavy use back in 1991, when Lynch borrowed 2,000 pounds (about $3,000) from an English pop promoter in a pub to start his first company, Cambridge Neurodynamics, from which Autonomy spun off in 1996. Lynch still sits on Neurodynamics' board, but he believes there is more scope in Autonomy, whose software and techniques he expects to be everywhere in a couple of years.

In a corporate context, you might use Autonomy's software to display links to material in the corporate archives that's relevant to a memo you're writing. On the Web, it might display links to news stories related to the ones you're reading.

Both companies are built upon the statistical principles derived by Thomas Bayes, an 18th century Presbyterian minister and mathematician whose formulations underlie the pattern-recognition algorithms that help computers behave more like humans, enabling them to recognize context and understand connections. Bayes' ideas on statistical models might never have been heard of had he not left his papers and 100 pounds to his friend and fellow clergyman Richard Price. Price published his late friend's "An Essay Towards Solving a Problem in the Doctrine of Chance," and promoted it by saying it proved the existence of God. In a computer context, however, Bayes' theorem pulls many probabilities together, enabling the computer to pick an answer out of multiple, often conflicting, pieces of information.

I have to ask because every mathematician I knew when I was briefly a math major worshipped him -- were you a fan of Martin Gardner's mathematical games column in Scientific American?

To be honest, no. But I came from an engineering background to math, and so it was very much driven by "What do I need to solve the problem?" rather than by delight in pure mathematics. But then, I would say that the area that I work in is pure philosophy anyway, because my area of research -- though I don't get to do pure research anymore, only what I think of as applied research -- is perception. Perception is very interesting because if you understand Bayes, if you start to think about it, you can see why Bayes is taught in philosophy classes these days.

I thought you were going to say he was taught in business schools.

Well, he ought to be, yes -- but keep that a secret.

The Free Software Project
Read Andrew Leonard's book-in-progress on Linux and open source -- and post your comments.