Signal's Meredith Whittaker says AI agents doing tasks on users' behalf pose security and privacy risks and refers to their use as “putting your brain in a jar”.
Signal's Meredith Whittaker says AI agents doing tasks on users' behalf pose security and privacy risks and refers to their use as “putting your brain in a jar”.
You're viewing a single thread.
This is the primary reason I've not given agents more power than something extremely controlled (I.e. only a function to turn on/off the lights). As I was always concerned that these generational models might accidentally do something dumb or annoying, let alone something that might be illegal or harmful.
I don't really see how anyone would ask AI to complete something completely autonomously at this stage, without oversight.
Good for image classification and tagging to automate sorting all the images off my phone. Some fun chatbots to soundboard ideas off. Anything involving credit cards, bank details, personal info near the internet? That's a big nope.