This paper shows, by discussing a number of Machine Learning (ML) appl
ications, that the existing ML techniques can be effectively applied i
n knowledge acquisition for expert systems, thereby alleviating the kn
own knowledge acquisition bottleneck. Analysis in domains of practical
interest indicates that the performance accuracy of knowledge induced
through learning from examples compares very favourably with the accu
racy of best human experts. Also, in addition to accuracy, there are e
ncouraging examples regarding the clarity and meaningfulness of induce
d knowledge. This points towards automated knowledge synthesis, althou
gh much further research is needed in this direction. The state of the
art of some approaches to Machine Learning is assessed relative to th
eir practical applicability and the characteristics of a problem domai
n.