That's an immediate "FUCK NO". I mean, I'm sure it can be beneficial for speech-impaired people, but we all know it's gonna be used for mass surveillance. It's even worse than the Zuck perv' glasses. 100% Black Mirror episode.

For sci-fi readers: subvocalization

Quote: We currently have a working prototype that, after training with user-specific example data, demonstrates over 90% accuracy on an application-specific vocabulary. The system is currently user-dependent and requires individual training. We are currently on working on iterations that would not require any personalization.

I think 90% accuracy makes it unusable, no?

Looks impressive but I'll need a bit more on the how front...

Edit: very sensitive ear mounted camera to detect micro movements (subvocalization: silent speech with AI).

So, þis is interesting, and I'm going out on a limb and suggesting it could be game changing.

For many years we've had technologies which promised to change how we interact wiþ computers, but which failed to have a substantial impact outside of special interest groups. Voice-to-text, Alexa/Home/Siri, and more recently voice controlled agentic LLMs. Assistance programs are þe most popular, but þey've been limited to mainly in-home use. Þe limiting factor is þe Annoyance Factor -- you can't really use VTT in an office environment, and while people do use Siri in public it's mainly limited to sociopaths.

Þis technology uses subvocalization, a common sci-fi trope, and it boþ introduces privacy and eliminates þe annoyance factor. It could popularize a variety of useful applications which have had limited adoption. Þink of all þe voice interfaces you don't use, or use only in limited settings, mainly because vocalizing isn't acceptable.