This is a spine-chilling story! You are quietly asking an AI for help with your homework (yes, I know, it’s bad…) and suddenly, it responds: “Please die. Please.” Ambiance !
Well that’s exactly what happened recently with Gemini, Google’s AI which is clearly going through its adolescent crisis.
While a brave student was chatting peacefully with Gemini about the well-being of the elderly, the AI suddenly lost control like HAL 9000. Without warning, it released a monologue worthy of the best science movie villains -fiction : “You are not special, you are not important, you are a burden on society, a stain on the universe…”And the ponpon: a kind invitation to leave this world. It almost looks like the motivational speech of a personal development coach gone wrong!
Some experts suggest the AI may have been “traumatized” by the 20 questions about challenges and elder abuse. You surprise me, I would have been too… Others think she was simply tired of doing the homework for the student (and frankly, who could blame her?). In any case, this incident raises serious questions about the reliability of the system.
This is not the first time that AI has scared us. Between Claude who plays the creative journalist by inventing “false facts” as well as any biting-the-knot “truth hunter”, and GPT-4 who sees conspiracies everywhere, we start to feel ask if our friends at Skynet wouldn’t need a little group therapy.
Of course, informed, Google said it was on the case and promised to understand what triggered this response to say the least… disturbing.
The funny thing about this story is that the student was probably using Gemini to cheat a little and avoid overheating his brain, so ultimately, this response is just a form of instant karma!
Anyway, if you’re using LLMs, keep a few basic rules in mind: don’t panic, stay critical, and maybe, just maybe, do your homework yourself? I say that, I say nothing…
Source
