The purpose of the TAOS project is to directly measure the number of Kuiper Belt Objects (KBOs) down to the typical size of cometary nuclei (a few km). In contrast to the direct detection of reflected light from a KBO by a large telescope where its brightness falls off roughly as the fourth power of its distance to the sun, an occultation survey relies on the light from the background stars thus is much less sensitive to that distance. The probability of such occultation events is so low that we will need to conduct 100 billion measurements per year in order to detect the ten to four thousand occultation events expected. Three small (20 inch), fast (f/1.9), wide-field (3 square degrees) robotic telescopes, equipped with a 2,048 × 2,048 CCD camera, are being deployed in central Taiwan. They will automatically monitor 3,000 stars every clear night for several years and operate in a coincidence mode so that the sequence and timing of a possible occultation event can be distinguished from false alarms. More telescopes on a north-south baseline so as to measure the size of an occultating KBO may be later added into the telescope array. We also anticipate a lot of byproducts on stellar astronomy based on the large amount (10,000 giga-bytes/year) of photometry data to be generated by TAOS.