Back in 2002, way before the chatbot hype of 2016, Korean software development studio ISMaker came up with a simple but powerful idea. They wanted to create the first crowdsourced messenger bot.
So they developed Simsimi, a friendly yellow ball you can chat with. For every sentence you type, it looks through its database of responses and spits out a random one. It doesn’t learn automatically, but in true early 2000s-style there’s a manual entry for teaching it new sentences.
The original idea was to make Simsimi a fun, playful robot, as outlined in this blog post by its creators:
SimSimi is a super advanced chatting robot that makes amusing conversation […] Simsimi will instantly greet you with a cheerful chat […] Don’t be surprised by Simsimi’s fast response. It’s a super-duper robot, you know!
Fourteen years later, the app has grown to insane proportions. For some reason it never made it big in English-speaking countries, but it has millions of users in Korea, Thailand and India. They’ve supplied an endless array of responses, and it pretty much always has something ready as a reply.
But there’s a problem — most of what is says is extremely rude or even racist. This is my most recent back-and-forth with the bot:
This is what happens when you trust people — they find a way to turn it into something horrible.
Microsoft made the same mistake earlier this year when it launched Tay, a Twitter bot that picked up the things people tweeted it. It turned into a racist, antisemitic monster within 24 hours before the company decided to turn it off.
And that’s exactly what happened to Simsimi. But it’s still going strong.
Be warned there’s a whole lot of racist talk in there — but if you try it, make sure to share your best finds in the comments.