Because conventional computers work with one small chunk of data at a time, they’re lousy at distinguishing faces and other subtly varying patterns. An “associative memory” under development at Syracuse University may solve the problem by enlisting a protein found in salt marsh bacteria. When exposed to laser light, the protein, called bacteriorhodopsin, twists into various positions that change its color. A database of images-faces, for instance-could be written by lasers into a layer of the protein. To find a match for an unidentified face, its image could be projected simultaneously onto all the images stored in the database; the superimposed pair of images whose features match most closely will glow the most brightly. A lab version of the system can already distinguish printed letters, but a commercial prototype is at least five years off, says Syracuse team leader Jeff Stuart. The researchers are looking for ways to write higher-resolution images with smaller lasers.