ニュース

The Economist2年

It doesn’t take much to make machine-learning algorithms go awry

This means modifying or adding extraneous information to a training data set so that an algorithm learns harmful or undesirable behaviours.