Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

> I can imagine this would be hard to "tout".

Open source client + whitepaper describing how the client encrypts the data with a particular scheme that still allows the server processing they need, without leaking undue data.

This can be as simple as: audio stream is discarded on device unless "ok Thing" is recognized, by a low quality open-source on-device recognition software. After that, the next 2 minutes of audio are sent to the mother-ship for higher-quality recognition and analysis.

Done, privacy-preserving Amazon Echo alternative. Get a third party (the EFF?) to audit it for you and put a badge that means to semi-technical users, 'this product goes beyond snake-oil on privacy'. Super-paranoid users can inspect the code for the client, which anyways includes little more than well-known open source libraries and some trivial glue code you don't care who copies anyway.

Of course, the real reason not to do this, is that companies don't want the 2 minutes of audio after the user asks their devices a question. They want the 'big data' of 24/7 surveillance (with all the beneficial applications this can have, but also the chilling ones).



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: