Researchers at Carnegie Mellon University are working on a silicon chip which would move automatic speech recognition from software into hardware.

“I can ask my cell phone to ‘Call Mom,'” says Rob A. Rutenbar who leads the team, “but I can’t dictate a detailed email complaint to my travel agent or navigate a complicated Internet database by voice alone.”

The problem is power. It takes a very powerful desktop computer to recognize arbitrary speech. Of course it would be impossible to put it inside a mobile phone, the batteries wouldn’t last.

The new silicon chip architecture would not only do speech recognition, but it would do it 100 to 1,000 times more efficiently than a conventional computer.

Researchers plan to unveil speech-recognition chip architecture in two to three years.

Press Release, via Kurzweil AI.