We may be looking at the 1GHz processor but the human brain is still better at data processing.

Not for long though. Two Korean-American scientists have developed a computer program that mimicks the way the human brain recognises patterns in images and text. They hope the program will improve Web searches and data processing by recognising recurring elements in an intelligent manner.

Daniel Lee, 29, and Sebastian Seung, 33, both researchers at Lucent's Bell Labs, tested the program with pictures of human faces and an encyclopaedia database. The program was able to recognise the main elements and then look for similarities, fitting the data into smaller sub-categories.

The technology could mean a vast improvement in Internet search engines, which currently work virtually blind to context. The latest search engine, Google, rates site importance according to how many other sites are linked to it. An intelligent engine would be an improvement on both these approaches.

Lee and Seung have a lot to learn about marketing, however. The report in which they announced the program was titled "Learning the Parts of Objects by Non-negative Matrix Factorization."

Now that's a catchy name... ®