Making AI Headlines · Page 6
Researchers Give Machines Better Hearing, Ending the Microphone's Easy Life
· New York time
Professor Paris Smaragdis is developing computers with superhuman hearing for uses including speech recognition, music understanding and environmental monitoring, according to MIT News. The work aims to make machines more capable of extracting useful information from sound. Listening is more complicated than turning up a microphone. Sounds arrive together, and the information a system needs may be mixed with other voices or background noise. Better methods for interpreting audio could help systems work with speech, music and environmental signals in situations where a simple recording is not enough. The research also offers a reminder that AI is not confined to the written answers of a chatbot. A listening system must deal with a continuous stream of sound and determine which parts matter for a task. Its success depends on more than being able to hear that something happened. There is a comic reversal in the ambition. People have spent generations telling machines to listen, and now the machines are being trained to do it more carefully than their owners. The microphone is no longer just expected to attend the meeting. It may be asked to sort out which conversation was the meeting, which was the music and which was the air conditioner contributing its usual unsolicited opinion.
Read this story in the full newspaper →Satire. Not factual reporting. Read the original reporting from the linked publishers.