liminal 🦠 on Nostr: The models trained to translate a thought "think about moving your finger", "think ...

The models trained to translate a thought "think about moving your finger", "think about biting into a lemon" into some action "click mouse", "rotate cube" are prediction models. "this eeg pattern is likely associated with this response", there is no need to decode the signals if all you care about is the correlations. Advertising machine learning models are a similar process "this type of advertising response is likely to produce this emotional output".

If this is an OS like Mac, Windows or Linux, I can understand that implicitly these models for training are not used for such actions or for the purposes of navigating the OS. But suppose that I'm browsing the web with this OS, is there a guarantee that that this raw data cannot be accessed by the host of websites?

Just to be very explicit here: this isn't an accusation of malice to the devs of this OS, rather its a question on the security and accessibility of this data to webservices/ third party software. But sure, I'd check out the papers