Week of October 3, 2005

Machine copies itself like DNA

A device that combines air hockey and DNA marks at least the third time this year the nascent field of self-replicating machines has seen a significant advance.

Researchers from the Massachusetts Institute of Technology have built a system that replicates strings of simple electronic devices from random parts floating on a cushion of air. The system works much like DNA, with a few rules and a few types of components working to make copies of an initial example for as long as the supply of parts lasts.

The electromechanical parts move around at random and attach or detatch from each other according to simple communications between neighboring parts. When a new part latches onto a part that has become part of a copy of the initial example string, the part in the string checks for a match.

The research could lead to self-replicating machines that build, repair, and -- reminiscent of childrens transformer toys -- reconfigure themselves for different tasks or changes in circumstances.

The MIT work is similar to a self-replicating machine developed at Cornell University, though the Cornell system uses carefully positioned rather than random parts. The MIT system is also similar to a simulation developed by researchers at the Canadian National Research Council and the University of Waterloo in Canada.

(Self-Replication from Random Parts, Nature, September 29, 2005)

Thoughts wander in VR

Devices that allow for direct brain-computer communications are advancing on two fronts: devices implanted in the brain, and external electrodes that pick up the brain's electrical signals. In general the external devices are safer and cheaper, but the implants provide more control.

Typically, users are able to control their brain waves in a way that moves a cursor up and down and side to side on a computer screen.

Researchers from Graz University of Technology in Austria, University College of London in England, Guger Technologies OEG in Austria and the University of Graz in Austria have expanded the amount of control available via external electrodes with a means of detecting changes in brain signals when users imagine moving their feet and translating the effect into walking motion within a virtual reality environment.

The system is distinguishes brain signals produced by thinking about different types of movements, including right hand, left hand, foot and tongue, and converts the information to control signals. Three test subjects were able to move forward through an immersive virtual street scene by imagining walking.

This type of interface promises to enable those who are disabled, and could eventually give and people whose hands and voices are otherwise occupied another way to assess computers.

(Walking from Thoughts: Not the Muscles Are Crucial, but the Brain Waves!, Presence 2005, London, September 21-23, 2005)

In a related development, members of the same research team have developed a portable brain-computer interface for use in virtual environments. The battery-powered interface, which includes a diskless Pocket PC, makes it possible for people to physically move while using brain signal control.

(Integration of a Brain-Computer Interface into Virtual Environments, Presence 2005, London, September 21-23, 2005)

DNA computing gets chipped

DNA computers can simultaneously check many possible answers to large problems like determining the best possible traveling salesman's route. There are billions of possible routes that include just 15 cities, and there are no mathematical shortcuts for solving this type of problem, which means they easily overwhelm ordinary computers.

Scientists are increasingly using chip-size devices that control the flow of DNA suspended in small amounts of fluids to carry out DNA computing; this is a key requirement for moving DNA computers from the chemistry lab to the computer science lab and eventually the data center.

Researchers from the University of California at Berkeley have built a prototype DNA microfluidic computer processor that brings DNA computing a step closer to practical use.

The processor contains DNA strands attached to magnetic beads that are suspended within the device by a magnetic field. Loose DNA strands that encode every possible answer to a particular problem circulate through the device, and complementary strands are captured by the suspended strands. Unattached strands are then rinsed away, and the cycle is repeated until only strands encoding the correct answer remain.

(An Integrated Microfluidic Processor for Single Nucleotide Polymorphism-Based DNA Computing, Lab on a Chip, October 2005)

We like people like us -- even fakes

Social science researchers have known for decades that we view people who mimic our body language more favorably than those who don't. It turns out that this effect holds true for our interactions with artificial people as well.

Stanford University researchers tested subjects who interacted with an artificial man or woman in an immersive virtual reality computer simulation. The agents either mimicked the subtle head movements of the test subject after a four second delay or displayed head movements recorded from a previous session with another subject.

The subjects rated the artificial agents that mimicked their own head movements as more persuasive and likable than the agents that did not. Subjects were also more likely to keep agents who mimic their head moments in view. (Eight out of 69 participants were judged to have consciously detected the mimicry, and their results were discounted.)

The work could be used to improve characters and avatars in video games and online environments.

(Digital Chameleons: Automatic Assimilation of Nonverbal Gestures in the Immersive Virtual Environments, Psychological Science, October 2005)

Bits and pieces: magnetic logic, finding the flu, mapping electrons

A logic scheme brings fast, cheap, low-power magnetic computer chips closer to reality (also see previous advance); a biochip that moves and heats fluids identifies a strain of flu virus; scientists map out how electrons move through individual molecules.