Hey folks, don't let me leave up one of those short, cryptic, off-the-cuff entries for two days. Yell at me in the guestbook, or something.

In the lobby of the building where my boss works, there's a "roboceptionist" - an information desk with a computer monitor showing a shoddy 3D rendering of a woman's head. You type in a question, and she talks to you. They've named her Valerie.

Not surprisingly, I've seen people act towards this thing in ways that they would never dream of acting towards a live receptionist. (I hope.) I've seen two young undergrads snickering as they type - gee, I wonder what they were typing? Today there was a middle-aged nerd standing in the middle of the hallway, just gawking at it.

I'm going to geek for a few minutes, so bear with me. This thing is a toy, no more. It doesn't represent any kind of useful innovation.

In thirty years, everyone who isn't a senior citizen will know how to use a mouse, and be familiar with the hypertext interaction paradigm. Contrary to what intuition might tell you, natural language is usually not the most efficient way to obtain information. The advantages of natural language are being able to ask vague questions, and being able to refine questions and get further information using linguistic shortcuts. But these are the kinds of questions that computers are very bad at answering.

Furthermore, there's an underlying assumption that people are the most comfortable interacting with other people. From personal experience I can tell you that if I need to ask a stupid question like "Where is room 3501?", or God forbid I get lost and have to come back and ask again, I'm going to be much more comfortable looking it up on something resembling a web page than asking a person. I don't think it's too much of a stretch that anyone who isn't totally uncomfortable using the web would agree with that.

Even granting the base assumption, it's gonna have to be a damned good system for people to interact with it in the same way they interact with a human. (There are people smarter than me who have been researching that kind of stuff for years, so I'm just armchairing it here.)

So if it's not more efficient, and it's not more comfortable, what's the point? It seems like the only people who would get excited about this sort of thing are the "OMG AI is so cool" crowd who never stop to consider that other solutions might be better, and the snickering teenagers telling poor Valerie to "SHOW ME YOUR TITS".

And, apparently, the people handing out the grant money.