Robot Receptionist with an Attitude
techno-vampire writes "Carnegie Mellon University is experimenting with a robot receptionist with a personality. The article on NPR tells about the receptionist, named Tank. Tank lives in a computer, with a Frankenstein-like face showing on the monitor. He responds to typed-in questions, including personal ones, with a rather curious personality courtesy of the Drama Department. Among other things, he doesn't seem to like his boss, Dr. Reid Simmons, very much. If asked, Tank will tell you he's also worked at NASA, and failed as a satellite robot. A job at the CIA was also a bust. Dr. Simmons explains that they're trying to make it easier for people to interact with robots, and upgrades are planned."
A good speech synth would add a lot to Tank's personality. (On the other hand, I have 1980s tech card that would sound awful but very robo-retro.)
One line blog. I hear that they're called Twitters now.
I think the point here is that the computer does not have the Stepford Wife annoyingly pleasant attitude that the usual computer assistants have.
Back in the early '80s my fellow students and I wrote computer based quizzing software for our classes. We played around with different responses to wrong answers. Contrary to what educational software companies were putting out, our programs would occasionally razz you for a wrong answer. Care to guess which ones the students used more often?
There is only so much a person can take of a caring and supportive computer before it gets really annoying.
BTW, I also wrote a rudimentary hash algorithm to weed out obscene names, without having to code those very names into the program. And yes, it could be defeated by inserting 1 or 0 in place of L or I and O.
So bring on the expert system AIs for Customer Service quickly please!!
He said expert systems. He didn't say replace customer service with text-to-speech ELIZAs. Give that guy a rough idea of how HTTP is supposed to work in training (which can be as simple as "client says GET webpage.html, server either says 200 OK and prints the page, or says 404 Not Found and prints an error page"), and when the customer says "telnet to your server", he can easily pull up a description of what Telnet is, an AI-influenced description of what he would need it for (by tracking the conversation): e.g., he'd need to know what port 80 means, but probably not what local echo means. Once he's connected, the system shows what the web page is supposed to return. If the customer says it's something different, the expert system has a link to the appropriate RFC, which he can check and either refute the customer or file a real problem report.
Most front-line customer service workers would never encounter telnet in their life. So we can't make knowing it a job requirement - but they have to know telnet if they're ever asked. So we give them an AI that they can use to learn stuff on the spot. It's a lot more helpful than making up stories or transferring calls all the time.