I hear two things requested a lot - an open source ChatGPT and more controls on the training/data/usage/etc.
How can these two things be held in balance? If it is open source entirely, can't I use the training data and tool in any way I want? What guardrails prevent me from using it to generate hate speech?
Given the compute power needed to train/run it; where are people going to do that in any accessible way that doesn't only allow people with a supercomputing cluster at home to use it?
How can these two things be held in balance? If it is open source entirely, can't I use the training data and tool in any way I want? What guardrails prevent me from using it to generate hate speech?
Given the compute power needed to train/run it; where are people going to do that in any accessible way that doesn't only allow people with a supercomputing cluster at home to use it?