Ashok Goel, a computer science professor at Georgia Institute of Technology, recently pulled a huge prank on his students: he gave them a robot for a teaching assistant. "Jill Watson" did regular TA work, like remind students of due dates and answer questions, all over email. According to a Wall Street Journal report, a few students were suspicious, but nobody uncovered the truth.

IBM wasn't directly involved in the Jill project — which is rare, because IBM is constantly looking out for new PR opportunities. The bot was trained by Georgia Tech researchers on 40,000 postings from a discussion forum, and only answers questions if "she" has a high confidence. And nobody figured it out!

So this is terrifying. Not because "Jill Watson" TA bots will rise up and kill us, but because if it's this easy to sneak AI into education-critical email conversations, it's going to start happening all over the place — especially where questions and answers are repetitive, and the field and scope are well defined. Also, professors like Goel with access to cutting-edge resources like Watson will get such good pranks it's not even fair. He plans to tell his students they have a computer TA next year, but won't tell them which one. You know, like with Cylons. This will end well.

Can we build a conscious computer?