Google’s AI hate speech detector is easily fooled by a few typos
Demand is growing for tools that automatically flag online hate speech, but the current crop can easily be fooled by adding a few typos or an innocuous word
Science and reality