i haven’t bothered to vet this
Google’s Gemini: “You Are A Drain On The Earth, A Blight On The Landscape… Please Die. Please.”
In a chilling episode in which artificial intelligence seemingly turned on its human master, Google’s Gemini AI chatbot coldly and emphatically told a Michigan college student that he is a “waste of time and resources” before instructing him to “please die.”
Given it’s ended up in a few outlets and some of note, and that it’s (sort of) verifiable it looks like there’s something there. Unsettling.
The unwritten assumption here is that machine learning should reflect the values of the person talking to it–and–that set of values should be the same for everybody.
Also, there is a technical problem where values are embedded in the training data. And, actually removing bias is not appealing to the people with the resources to do it. Plus, it might not be wise to fix text prediction by erasing our history.
I think it is a mistake to expect compassion from AI. And people interacting with it should expect the full range of human expression: that which was input into the models and forms an inescapable part of it.