To elaborate on what TrueTears has said, basically the fundamental basis of AAS is that metals will absorb light of a particular wavelength, and that the amount transmitted or absorbed, would be proportional to the concentration of that metal in the sample. However, do remember that when these metals absorb this light, they're in an unstable state and will eventually emit this exact same wavelength of light again to reach a stable 'ground' state again. This causes problems for the AAS technique as you're essentially having the light which was previously absorbed being re-emitted and reaching the detector. This would cause the detector to register this light as well, and would cause the detector to think that the amount of light transmitted is greater than it actually is. Hence, the light from the lamp is pulsed, and the detector is designed to only count light which is pulsed to avoid confusing light from the lamp, and light which is re-emitted.
EDIT: TT beat me to it D: