Thread · 6 tweets · 21 Sept 2021

"It's okay because it's on-device" is one of the coming biggest issues in tech. Sometimes it's true, often it's a lie. (Arguably unintentional, but that hardly matters.) A few quick thoughts: 🧵

Who does your device/OS/browser work for? That device might feel more trustworthy because you bought and therefore own it, but without fiduciary obligations there is no guarantee that it will put your interest before its maker's. This needs rules. papers.ssrn.com/sol3/papers.cf…

It doesn't magic privacy issues away. Take the classic case of Target sending a young woman pregnancy-based ads before she was ready to share with family. Is the same violation computed on-device with on-device ads in any way different?

More generally, is carrying out inappropriate processing on the device better than off device? Is using one's control of the device or agent to monetise another business's work no longer theft if it happens on device?

Questions of whether the user retains agency, of the application of persuasive tech, of control over parameters, of the uniformisation of collective cognition and the organisation of information all persist — often in exactly the same shape.

The gain from moving processing on-device is largely from a security perspective. Definitely a plus, but hardly the primary threat model. And… that's about it?