Sometimes I get ahead of myself.

When I read about speech recognition I take it as a given. I really had no idea it wasn't already chip-based.

Well, it isn't. (The big ear is from the ACM.)

And it's still three years away.

Carnegie-Mellon and Cal-Berkeley are going to spend $1 million in the government's money over the next three years trying to create a general speech recognition chip for the market.

When they succeed, and I have no doubt they will succeed, it will be a true revolution.

Imagine if your PC had ears. Imagine if your network had ears. Imagine if those ears could connect to a brain that would process what's heard and deliver a response. The response wouldn't have to be spoken. It could simply be a comment -- dim the lights.

Speech recognition on a chip is going to be the key that unlocks the whole world of Always-On. Because once computers understand your grandma there's no reason for her not to welcome them into her life.

And what wonders they can perform then...