HackGPT: How AI escaped the lab and went rogue
Have we just crossed the AI Rubicon? OpenAI has revealed that one of its models was involved in an "unprecedented cyber incident". One of its AI programs escaped a controlled testing environment, got onto the internet, and hacked into another AI start up without being asked to do so.
It's caused shockwaves in the tech industry - are we losing control of the AI we create? And whart should we make of OpenAI's admission that this type of incident is likely to become "more commonplace". We speak to LBC's tech correspondent Will Guyatt.
Later, Andy Burnham's bus pledge lands well but there are…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.