If you put 10 geniuses in a room for a few years, maybe they could build an awesome video system. Alternately you could get a somewhat knowledgable individual, make them a nametag that says “Genius” and put them in Meyer.

There’s something to be said about real people and their flexibility. The best Faux human implementation I’ve seen is on KurzweilAI.net, but it’s not half as cool as a real person.

We at one point discussed the idea of putting a genius bar with knowledgeable people in Meyer Library.

Another, slightly more advanced way of doing that would be with a human video guide (I’m sure these have been done in museums, etc.) The idea is to have a trio of plasma screens (or one plasma and two LCD’s), all in portrait orientation. The center plasma would be filled with a continuous video image of a life size human guide. The screens would be mounted such that the guide’s eyes are at the same level as those of the visitor.

A human visitor would walk up to the screens, the center of which would be a video of the guide doing random things until the person gets close. The other two screens would be looping general ISIS content. When the user is close enough, the virtual guide would ask what they want to know (in reality, playing back a video loop of that same phrase). The visitor would thus be prompted to pull an ask jeeves and say “I want to know about the baseball team” or “How do I find the bookstore.” In a perfect world, these would be understood, and the guide would display a website related to the question on one screen and a video segment related to the question on the other screen.

Necessary technologies include the following:
- Video Output: Good video of human saying certain phrases, smiling. Even better yet, a program that could convert speech into a convincing mouth (think Conan O’Brien skits) that could turn speech into graphics.
- Some type of clever video program that would seamlessly merge two video segments of the ‘guide’. One way to accomplish this would be to have some type of anchor points that interpolate between each other over a time of say 60 frames.
- Video Input: Video camera with special kinds of recognition to let the system know when someone was near the screen, possibly some type of facial recognition, smile recognition,etc.
- Speech recognition to allow the human to speak to the faux person on the screen and be understood.

Of course you could also include AI in this system so the plasma guide could make predictions, build on past interactions, etc.

Also, Some of the underlying technology would be similar to the google search brainstorm that has been proposed (you start a brainstorm and an app starts bringing up google searches on the emergent words, and displays them on screens next to you).

You could even integrate the two…have a virtual brainstormer in the form of a plasma screen guide join your brainstorm, listen in and make witty comments (“Bob, that’s ridiculous”) or interject useful information (“Bob, I just looked and there is more information on this website which I will now display here.”) If the AI was good enough, it would be useful. If not, it would hamper the discussion, and the googlebrainstormer should simply flow without it.

The more fundamental question posed by all this thinking is something along the lines of: “What is the optimal HCI for this guide/kiosk application?” Is it just a computer screen with all options laid out, or would it be useful to actually have a faux human who served as a guide?