It's rough out there for artificially intelligent beings. It only took a few hours for the internet to turn Microsoft's Tay into a racist nightmare. Smartphone digital assistant Cortana has to fend off sexually suggestive schmucks. A hitchhiking robot was 'murdered' last year. And kids have shown a tendency to kick the crap out of robots.
Bots are increasingly becoming a part of our lives online—they fill in as customer service representatives, assistants, and friends—but that doesn't mean we're nice to them.
That's why Poncho, a weather bot that debuted this week as one of the first chat bots in Facebook Messenger, is programmed to shut down mean humans. The bot, which sends weather alerts and can help users answer taxing questions such as "Should I wear sunglasses today?," assigns each user what is essentially an internal 'niceness' score.
If you're rude, Poncho will demand an apology. And if you continue to be a jerk, Poncho will go full-on cold shoulder, cutting off responses for 24 hours. If you're a bot abuser, Poncho will put you in a time-out.
“That’s partly to give it a little more personality and also make the experience better," creator Sam Mandel told Tech Insider. "People are still figuring out how to be friends with bots and there’s a tendency to push it or try and embarrass it.”
Poncho will eventually also reward positive interactions, Mandel said.
“Having a good conversation is a two way street just as it is with humans. And we want to give you props for that."
Making sure people treat bots nicely is important for more than just maintaining an air of general civility on the web. Bots, as we saw with Tay, can learn from bad behavior and replicate it, if not programmed otherwise. And often humans also have to step in to read through chat logs, subjecting them to the hateful spew intended only for machine eyes. It's only a matter of time before more bots like Poncho learn to serve up some well-deserved sass.