Hartmut Neven, who does research in computer vision at USC, dreams of using cameraphones to build a database that would essentially Google the world.

Take a picture of a painting and get an audio recording, through your phone, all about it. Take a picture of a restaurant and get a review.

There's even a sound business model for such a plan. But, as Google's local service illustrates, there's also one big hurdle toward creating it.

Interactivity.

If you're to build this kind of database, you must have the capability to take input directly from users and put it into the database.

Google has yet to prove itself truly capable of this.

It can make a start by improving interaction with Google Local, and by advertising that interaction.

For instance, this list of sports bars supposedly near my house is completely obsolete. I know of at least two within a mile of here that aren't listed, and I know of two more that have come-and-gone within that space in the last year.

There's also the problem of ambiguity. Input the word bar in Google local, and you don't get any service. You get some lawyers, a Thai restaurant, and a closed antiques store. The Yellow Pages can do better, much better, and they come out once a year.

Rather than building a great picture database, get interactivity right. And, while you're doing that, allow business owners to add to their listings with pictures. Kill two birds with one stone.