Cutting edge: Digital Darwinism

February 1, 2002

The future of computers is not artificial intelligence, says Peter Bentley, but true intelligence, in the form of software based on human biology.

Children's toys are extraordinary these days. Little robot dogs that can learn to beg, roll over, be "happy" or "sad". But then adults' toys are equally amazing. Mobile phones that understand your voice. Cars that can work out where they are and talk to the driver. And when you use your credit card, intelligent fraud-detection systems monitor and learn your normal patterns of purchases. Should someone steal the card and start buying something you wouldn't, the clever software notices.

The way these technologies work is surprising. Neural networks modelled on human neurons, but made from software, enable the dogs to learn, remember and make simple decisions. Similar software listens to and understands your voice when you talk into your phone, or learns your patterns of purchases with your credit card.

We also make use of evolution: with solutions to problems reproducing inside the digital universe of the computer. These children inherit digital genes from their parents, and should those genes make them into good or better solutions, they will go on to have offspring themselves. Generation by generation, our computers evolve solutions to problems.

I have been working with biologists to understand and exploit various processes of nature in our technology. I began my research in the field of evolutionary design. My PhD research showed that it was possible for a computer to evolve designs from scratch. Aerodynamic car and train shapes, hydrodynamic boat hulls, even furniture was evolved by my computer. I am known for my evolved coffee table. I did not tell the computer what the designs should be, I set up a digital environment in which designs had to perform the right function in order to survive and reproduce. The designs that emerged showed that evolution in my computer was better at designing than me. I went on to use evolution to compose music and to learn how to detect fraud, this time in collaboration with Lloyds TSB.

These days, my research covers more unlikely areas. For example, I am working with PhD student Jungwon Kim to create a computational immune system that lives in a computer network waiting to pounce on hackers or viruses. This uses many of the processes from our own immune system, such as ideas of B-cells and antibodies.

I am working with PhD student Sanjeev Kumar on an investigation of computational development. We have modelled developmental systems to help biologists understand how they work, and we are creating a digital developmental system that will enable designs to grow from evolved digital genes using proteins and cells.

I am working with ecologist Jacqui Dyer to model the evolution of plants in disturbed environments. We are able to demonstrate how plant life has evolved to cope with or even utilise disasters such as fires or droughts - all inside a computer.

My research with PhD student Tim Gordon investigates how computers can evolve novel electronic circuit designs directly into silicon, using special chips called FPGAs. Supiya Ujjin and I are examining how evolution and digital insect swarms can learn human preferences, enabling a computer to help customers find what they really want to buy in internet shops. And my research with Tim Blackwell is demonstrating that digital swarms and flocks can improvise music in real time, as a musician plays.

Digital biology is changing the way we use computers. It is also transforming the capabilities of computers. Artificial Intelligence was the title of a recent movie - truly intelligent computers using digital biology are nearly a reality.

Peter J. Bentley is an honorary research fellow in the department of computer science at University College London and author of Digital Biology (Hodder Headline 2001).

Please login or register to read this article.

Register to continue

Get a month's unlimited access to THE content online. Just register and complete your career summary.

Registration is free and only takes a moment. Once registered you can read a total of 3 articles each month, plus:

  • Sign up for the editor's highlights
  • Receive World University Rankings news first
  • Get job alerts, shortlist jobs and save job searches
  • Participate in reader discussions and post comments

Have your say

Log in or register to post comments