Journalism Without Journalists

Date: Fri, 21 Jul 1995 20:11:29 -0400 (EDT)
Subject: Cat people
X-UIDL: 805309723.000


I would have to say that I am a "cat person" too. I do not want to have to spend a lot of time sitting 18" away from my monitor typing away at a keyboard and trying to find my mouse.

When I foresee people personalizing their computers in the near future, I do not see them typing in arcane commands. I do not see them being fixed in front of a monitor, unable to walk around.

What we have been working on, ever since I began the Architecture Machine Group, is a computer that is more of a companion than a glorified appliance. I want a computer that I can talk to, and not necessarily from the same room. I want a computer that will understand my facial expressions and my tone of voice. Even a dog (or cat) can do that.

We have been working on voice recognition (and the recognition of facial expressions and gestures) at The Media Lab. As I wrote in Being Digital:

The problem of speech recognition has three variables: vocabulary size, degree of speaker independence, and word connectedness, the extent to which words can be slurred together as they are in the cadence of normal human speech. (Negroponte, 1995, p. 140).

Each one of these variables can be thought of as an axis on a three dimensional cube. Where the axes meet, in one corner, there is no speech recognition at all. Vocabulary size equals 0, as does the degree of speaker independence and the understanding of connected words.

At the opposite corner, there is perfect speech recognition. While it may be impossible to reach that corner, I think we already can find a place somewhere in between.

The problem posed by each of the speech recognition variables can be overcome. Instead of having a computer be able to recognize 50,000 words at any one time, we can save space, memory and computing power by having it be able to recognize only 500 words at one time. Those words could be linked to stored words that would not have to be resident in the computer's memory. For instance, if I told my computer to place a phone call, my Rolodex could be loaded.

The other two variables can be dealt with just as creatively. I expect that we will be talking with computers quite a bit in the near future.


image map

Back to beginning|Previous message|Next message|References

© 1995, Troy Wolverton. Disclaimer

Back to my home page