This proof of concept device is exactly the kind of direction I've been hoping that personal electronics would take: shifting from the consumption of media (tapes, CDs, books, etc.) to the creation of it (cameras, small writing devices, microphones).

In a nutshell, this device is constantly recording the sound around you, via a microphone at your waist and one in your ear and, upon recognizing a preprogrammed keyword or phrase, like "nice to meet you" or "I'll have to remember that", will store a predetermined portion of that buffer for you, for further retrieval.

They give the 'cocktail party' example of remembering the names of people you met (which I think would only be useful if there was also an integrated camera) but I could see countless uses in a classroom environement or social environment; anywhere you might hear something you want to remember, but weren't prepared for.

With the proper software, this could also be a very useful tool for people with ADD, giving them the ability to, with a simple voice or button-driven command, get a playback of the last 30 seconds of conversation, if their mind wandered. In fact, for some ADD people, the ability to run a constant, lower volume stream of the conversation, delayed 5-30 seconds from 'realtime' could help them keep their mind in the conversation, even as it wanders, with the first pass being a 'screening pass' and the second enabling them to turn their focus on the conversation.

The device also has a clock and GPS, so with each saved annotation would be the time and exact place where the conversation took place.

There aren't any commercial production plans yet, but this kind of technology could easily be added to existing consumer devices, such as cellphones or HD MP3 players, given the proper firmware and some hardware tweaks.