51–60 of 295 posts
I was never aware of this, and while poking around I came across ~/Library/Trial/Treatments/ /factorPacks/ /assets/com.apple.siri.asr.dictation.*/etiquette.json These files do contain an impressive amount of swear words, some of which I had never heard.
Open source is where you don't have to apply guesswork to find out what a suspicious looking piece of software is doing on your computer.
Maybe some user can be unaware of what is happening with OSS, but the community as a whole will always have accurate information of what is what (unlike this case where everyone is guessing).
Open source is where you don't have to apply guesswork to find out what a suspicious looking piece of software is doing on your computer.
Nope, instead you have to apply guesswork to figure out why an apt-get call accidentally uninstalled your desktop environment. :P
Earlier quoted context omitted.
Of course not?
Why? (Assumimg Vscode respects the setting)
If you found that your next door neighbor was sneaking into your house while you're away, and he said "oh I have an option not to do that. Just ask me not to any more." Would that be good enough? You say "ok I would like to invoke that option" and that's it, you happily go away on a month long vacation without installing any cameras or anything because they guy said he won't do that now.
That's a single package, whose setting affects no other apps, and you have to trust that the app even really does what it says, which makes no sense when the topic is apps that may be doing unwanted things behind your back.
I trust that Apple isn't doing anything nefarious here, but at the same time, I'm quite grateful that others aren't so trusting. I want the "white hat" suspicious and skeptical rooting through everything. In fact, I don't think I'd trust Apple so much if it weren't for those that don't trust them. Their constructive distrust enables my trust.
I just looked into ~/Library/Trial on my main machine (macOS 12.2.1). I got 380MB spread over 584 files (while TFA's author got 300k???). Most if not all of them are related to com.apple.siri.asr.dictation.en_US. Close to 100% of the space is taken up by .fst files, e.g. $ file /path/to/msg-bigG.squeezed_quantized_acceptor.fst /path/to/msg-bigG.squeezed_quantized_acceptor.fst: OpenFst binary FST data, fst type: squee…
I checked mine, and I see lots of mention of "Siri" (TEXT TO SPEECH, FIND MY, etc). Seems most likely this is A/B testing for ML models, such as those for Siri and Photos analysis? That's my best guess. Since Siri and Photos now processes more locally (especially on M1's dedicated ML cores), all the training improvements and testing probably happen locally as well. It makes sense it would regularly download updates f…
The privacy implications of this are dubious. There is as-yet no good research showing how these finetuned models can be used to reveal exactly what photos you have in your photo library, but many experts believe that won't be the case for long.
I just looked into ~/Library/Trial on my main machine (macOS 12.2.1). I got 380MB spread over 584 files (while TFA's author got 300k???). Most if not all of them are related to com.apple.siri.asr.dictation.en_US. Close to 100% of the space is taken up by .fst files, e.g. $ file /path/to/msg-bigG.squeezed_quantized_acceptor.fst /path/to/msg-bigG.squeezed_quantized_acceptor.fst: OpenFst binary FST data, fst type: squee…
I looked at one of the files: `~/Library/Trial/Treatments/SIRI_DIALOG_ASSETS/factorPacks/6242008c0f185b3e2ef6bf8e/assets/com.apple.siri.dialog.socialconversationfl#d5b431b/dialog/SocialConversation.catfamily/dalWhoIsYourValentine.cat/en.cat.bin` the file is gzip compressed and the decompressed data reveals some strings that are probably related to the question asked in the `cat` file: tag:valentine";Only localize if…
Would be much nicer if they stated the purpose and allowed for some configuration, though. And verification by third parties is obviously still important.