I'm referring to Cyc (as in 'sike'), a system quietly released in the summer by Cycorp (www.cyc.com) after no less than 17 years' development.

Since 1984, former Stanford professor and AI luminary of the 1970s and 1980s Doug Lenat has been sitting in Austin, Texas, compiling a database of 1.5 million facts about the world to help give computers 'common sense'.

Assumptions underlie all statements, it turns out. Lenat is reported to have started his project with the following two sentences, which he has said (in a modern, text-based version of the famous Turing Test) the computer must understand to be deemed 'intelligent': 'Napoleon died in 1821. Wellington was greatly saddened.' Any real AI must not only know who these two historical figures were, but also what death, life, sadness... are.

Lenat is a controversial figure in a controversial field. He's been squirreled away working on Cyc for many years, rousing the ire of his more mainstream AI colleagues by simultaneously refusing to publish any provisional findings in any respected academic journals while at the same time criticising former comrades for having given up on the whole AI project - to create a true artificial mind.

So what might Lenat and his team have achieved? The nature, limitations, and structure of common sense reasoning is itself a massive philosophical project. What makes Lenat so sure he's cracked it?

Well, the good news is you can check it out for yourself as Lenat is making an open source version of Cyc available for you to test-run.

The company ultimately intends to sell Cyc, and one can see the possible applications in a number of information management and search applications.

Ad hoc ramblings or true machine intelligence breakthrough? You can help decide.