Hacker News new | past | comments | ask | show | jobs | submit login
[dupe] Facebook collected and transcribed users’ audio without permission (techcrunch.com)
24 points by sahin-boydas 11 days ago | hide | past | web | favorite | 9 comments

This title isn't great. Users knew their audio was being collected and transcribed. The issue is that users (might have) thought that the transcription was automated vs being done by real humans.

Apparently FB, at least in some instances, had people do the transcriptions (possibly just to double check the automated transcriptions).

Unless FB's speech recognition has made award-winning improvements, humans are double-checking virtually all of it. Even so, I'm skeptical that FB's wording of any notice was clear, whatever the arrangement.

According to Bloomberg the transcripts were opt-in:

> The company said the users who were affected chose the option in Facebook’s Messenger app to have their voice chats transcribed. The contractors were checking whether Facebook’s artificial intelligence correctly interpreted the messages, which were anonymized.

According to my experience they aren't. I sent a massive burp to the group chat, saw some new "transcribe" option under the audio file, which I ignored. Later I checked the chat and my buddy had "enabled transcriptions for this chat" and my audio file had "oh my god," I guess its transcription of my burp, as a caption.

The "data anonymization" defense is a pet peeve of mine and it's truly meaningless in this case. Stripping out explicitly identifying information from the data doesn't provide much privacy when the rest of the data potentially consist of things like your mom reciting her SSN to you or you discussing your doctors appointments in detail. Even if they vetted only seemingly trivial transcriptions, there is the possibility that, e.g., the computer heard "my chaise couch" when a human would clearly hear "my Chase account."

I don't know too much about this so apologies if this misses the mark, but how much of this was due to it being opt out? Or hidden behind confusing options?

The article explicitly says it was opt-in. How could that be "due to it being opt out"?

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