Galaxy Zoo the Right Approach?

I've been pondering the premise of Galaxy Zoo, a newly launched astronomy Web site that lets the public contribute to a project that aims to help astronomers understand the large-scale structure of the universe. It adopts the basic model of SETI@home in enlisting volunteers to help analyze huge amounts of data in the search for extraterrestrial intelligence. Today there are more than 85,000 volunteers.

In Galaxy Zoo, participants are called upon to help classify a million or so galaxies based on images taken by the Sloan Digital Sky Survey (SDSS). The problem is framed in such manner that participants simply discriminate between a spiral or elliptical shape. But whereas SETI@home is a grid computing project in which computers do the number crunching, Galaxy Zoo participants must devote their eyes and minds to the task of evaluating the galaxies­, pattern-recognition work that (so the organisers believe) people do much better than computers.

Is it true that people are much better than computers at such recognition tasks? Today I had the honor of recording a lecture by Dr. Stanley Osher, an American mathematician who specializes in image recognition. Dr. Osher developed the Level Set method for tracking images and shapes in motion, now widely used in crime detection and animation. After the lecture I asked him about the Galaxy Zoo problem and he assured me that it was a waste of human effort - that computers could do the job more efficiently and accurately.

I buy the basic premise that humans have very well-refined pattern recognition capabilities, but I think the Galaxy Zoo problem is so basic that a machine could easily handle it. This is especially true if the software is as powerful as the various examples Dr. Osher shared in his presentation. The organisers seem to think otherwise, but it is indeed wonderful that they are inspiring so much popular interest in astronomy.