Nature can improve our computers

| Friday August 27, 2010 00:00

Robots on Mars that can fix themselves and computers built from DNA: not science fiction but the work of scientists at the forefront of computing.

Dennis Shasha, 55, is a professor of computer science at New York University's Courant Institute of Mathematical Sciences and the author, with Cathy Lazere, of Natural Computing: DNA, Quantum Bits and the Future of Smart Machines (Norton), a survey of research in fields as disparate as engineering and medicine. This New Yorker sees an emerging common theme: that the future of computing lies in a synthesis with nature.

What can computers learn from biology?We should look at the history of the two fields and how surprising it is that they should come together. Computing was really born of physics and of this 'clean-room' mentality - and a lot of computing is still like that.

When one tries to control every bit of accuracy and tries to ensure that nothing could possibly go wrong - well, this is very different from the messiness of nature.

But as computers become more mobile and autonomous - either they can gain their own power from the environment or they have a long battery or are wireless - their problems become quite different.

Instead of problems being algorithmic, which means they can be expressed as a recipe - you do one step and then the next and you finally arrive at an answer - now their problem can be: how do I survive? If the problem is how do I survive, then all of the survival mechanisms that organic nature uses become relevant.

And robots that can repair themselves are an example of this?In the history of space travel, it's been easier to build a computer program to guide a spaceship to Mars than it has been to build a robot that can navigate the terrain there with anything like the skill of goat.

A professor of robotics at the Massachusetts Institute of Technology (MIT) called Rodney Brooks saw a problem with what's known as the 'monarch' model: sensors provide data to some kind of higher intelligence, which in turn figures out a model of the world before sending a set of decisions to a set of actuators (wheels, for instance).

But it's very hard to construct a model of the real world. Brooks thought that the real world is its own best model, and that it's much better to have very low-level intelligences, which can work together.

One intelligence may say: 'Do not collide with anything' - something fundamental to avoid a bad outcome - whereas another one may say: 'Explore as much as possible.' But if that one conflicts with the first one, than the first one wins. And so on.

As a result, in the 80s, Brooks was able to construct very simple robots that didn't need much computing power but which were able to achieve things that other robots couldn't.

He suggested sending lots of little robots to Mars, which should just be let go, as opposed to highly controlled robots, which would be very expensive, could easily break and would in fact be very hard to control over all that distance.

It proved a very influential idea and other scientists are now thinking about how robots might adopt a random - almost an evolutionary - way of fixing themselves using what are known as genetic algorithms.

When we think of computers, we think of electronic devices such as laptops, but is there a role for nature here? A chemist called Ned Seeman, a colleague of mine at New York University, came up with an entirely new way of controlling nature in the early 90s when he realised he could force DNA to self-assemble into what might be thought of as sculptures.

DNA is made up of the bases A, C, G and T - and the A likes to bind with the T, and the C likes to bind with the G. If you can make single strands of DNA in the proper way, you can make them form stars and dodecahedra and all kinds of shapes.

Lately, Ned and others have been building robots out of DNA - tiny robots, but robots in the sense that one can control their movement.

There are millions of them, so it is conceivable to have them detect the DNA of some bacteria and either light up a signal that says: 'Here's these bacteria and you have to apply those antibiotics' or maybe even conceivably fight them, using other enzymes. This is all in the future, but it's not in the very far future.

What it consists of is thinking in a computational way - giving instructions to something. Computer scientists are often on a kind of power trip - telling the computer to do this and then that.

And in this case biologists are thinking computationally - urging DNA to do this, that and then the other thing. It's a very similar way of thinking. Fans of this synthesis between biology and computing have been around for some time, but it's only now that they are coming together in different fields.(The Guardian)