The goal of this study was to analyze the magnetoencephalogram (MEG) background activity in patients with Alzheimer's disease (AD) using the auto mutual information (AMI). Applied to time series, AMI provides a measure of future points predictability from past points. Five minutes of recording were acquired with a 148-channel whole-head magnetometer (MAGNES 2500 WH, 4D neuroimaging) in 12 patients with probable AD and 12 elderly control subjects. Artifact-free epochs of 20 seconds (3392 points, sample frequency of 169.6 Hz) were selected for our study. Our results showed that the absolute values of the averaged decline rate of AMI were lower in AD patients than in control subjects for all channels. In addition, there were statistically significant differences (p<0.01, Student's t-test) in most channels. These preliminary results suggest that neuronal dysfunction in AD is associated with differences in the dynamical processes underlying the MEG recording.