405

When data training goes wrong (lemmy.world)

submitted 7 months ago by devilish666@lemmy.world to c/programmer_humor@programming.dev

30 comments fedilink hide all child comments

you are viewing a single comment's thread
view the rest of the comments

[-] pineapplelover@lemm.ee 23 points 7 months ago

Sound recognition? Is this an always on thing in iphones. I hope it doesn't get sent back to servers or something

[-] HEXN3T@lemmy.blahaj.zone 38 points 7 months ago

It's an accessibility option for the hearing impaired, e.g. doorbells, knocking, crying, smoke detector alarms, and some others.

[-] pineapplelover@lemm.ee 7 points 7 months ago

That's pretty cool. Iirc a lot of disabled people like using iphones for this reason. I hope we can have better accessibility tools on the Android side soon.

[-] ji17br@lemmy.ml 12 points 7 months ago

It’s all done on-device

[+] vox@sopuli.xyz -34 points 7 months ago* (last edited 7 months ago)

(also I'm fvcking tired of privacy extremists on here. if something is proprietary it doesn't mean that it uploads literally everything to the server, on-device processing is still a thing y'know)

[-] DreadPotato@sopuli.xyz 17 points 7 months ago* (last edited 7 months ago)

on-device processing is still a thing

Unfortunately even if this is initially processed locally, there's a significant risk that results and/or samples are still sent back. manufacturers have shown on multiple occasions that the allure of data they can leverage is too much to resist.

[-] pineapplelover@lemm.ee 4 points 7 months ago

Somebody mentioned their Google Takeout (export of Google data) contained audio of them using the Google assistant. So Apple could technically do the same thing.

I'm not a privacy extremist, I just want to understand how technology works and how you should be using things in order to be safe online.