I think Apple is slicing the baloney pretty thin when it says audio isnā€™t recorded. Surely thereā€™s some recording because summaries require context; they canā€™t be made on the fly. Maybe a transcription is done in near-real time and the summary is drawn from that. If so, the transcription is a type of recording, albeit one that isnā€™t stored after the summary is made. As for not identifying the speakers, I donā€™t see how that matters. Old-fashioned audio recordings donā€™t identify speakers, either, but theyā€™re still intrusions on the speakersā€™ privacy if they didnā€™t give consent.

Apple has always focused on the privacy of its customers: your information is encrypted, your data stays on your device. Itā€™s following that same pattern here, but the privacy concerns of Live Rewind and Siri Recap arenā€™t limited to the owners of Apple Watches. Apple now needs think about the privacy of non-customers and be able to explain how thatā€™s being protected, too.

And while an argument can be made that textual transcriptions and recaps arenā€™t as invasive as saved audio recordings, that doesnā€™t mean they arenā€™t invasive at all. That little lock animation Appleā€™s been using to describe its commitment to privacy has been very effective, but its binary nature doesnā€™t fit when the products start putting privacy on a sliding scale.

Source: New Apple Watches have ears

I think this is a bigger deal than that. Apple just demonstrated that 24/7 live recording and transcription is possible. Guess who cares about the transcription and summary even more than Apple - Meta and Google and OpenAI and Anthropic - companies that are willing to perennially record you and glean information about what you speak and care about.

If it can be done on a device the size of the apple watch, you can bet it can be done on a phone and then on home power connected devices.

You know that myth where everybody thinks that instagram is always listening to you - buddy that just became real.

Apple’s leading with a potential value for the consumer here - maybe. As someone that used granola in previous corporate settings, I want to assuage everyone that a summarized notes from every meeting is NO MORE useful than having written notes. Finally, it also doesn’t necessarily make any agent better by having that information. There’s something called as relevant even within context and increasing the context does NOT improve its relevance.

I might be being cynical - however - game theory predicts that a move like this by Apple just gave permission to a lot of actors to take this to the next level.

Amazon, Google and whatever OpenAI is cooking up are all going to enable permanent listening on their home devices.

While mobile computers are still not constantly listening to us, we are probably only one to two generations away where the hardware is going to be part of all those devices and then move to a default-on state soon after.

Meta is going to turn this on for all their Meta glasses to match the capability that Google and Amazon, their competing ad platforms will develop as a result.

Humanity is going to enter a world where every sound is going to be recorded and synthesized and summarized. I know people are predicting that current regulation could help us prevent from entering this world. My prediction is likely that the regulation is going to change to accommodate for the reality we are going to find ourselves in.

The weird times are just starting.