In the early days of AI, it was noted that chatbots fed with data scraped from the internet tended to go off on bizarre tangents amplifying racist, conspiracist, and Nazi ideologies.
Some work was done to reduce or eliminate this problem. But then again, there are now people in power who see it as a feature, not a problem.
With #Twitter scraping data from all of its existing users' posts–a userbase more and more tilted to the alt-right–it's easy to imagine what its AI is going to spew out.
@DavidSalo Blenderbot! He used to spew racist stuff and praise Zuck. I tried toying with him by giving him one of those moral questions that goes in circles with no solution. He got frustrated and did an AI mic drop. I also suggested his master is evil and he'd agree before singing his praises. I used to spank him when ever he spewed racist statements about things like slavery.