[epberry]

I’ve long been interested in the “cocktail party problem” which involves disambiguating audio in a conversation with multiple people. I think this tech is foundational for better video calls and smart speaker devices for homes. The best research I’ve seen on this is from Mitsubishi but as far as I know this is well into the territory of an impossible problem today.

[nappy]

There's been a ton of interesting research and progress on this in the last few years: https://ai.googleblog.com/2018/04/looking-to-listen-audio-vi...