Apparently this algorithm can be used to modify and corrupt the machine learning processes of artificial intelligence. Big companies are already using AI for a variety of things, so a corruption could hurt a lot of people and cost a lot of money.
But this is not actually the biggest problem with AI. Turns out that if the programmer is racist or biases, the AI will be trained to be racist or biased.
And if the owner of the AI happens to be an intrusive asshole, like Facebook or other social networks, then guess what? The AI will be intrusive, assshole-ish and will be ten times worse than people can ever be.