If most of the internet is nonsense, the output of most AI will be nonsense.
We're quickly going to hit a point where a significant amount of the input into LLMs will be generated by LLMs. That seems like a dangerous precipice — assuming people continue to take this tech over-seriously.
@davetroy beyond the intellectual property questions, I've been wondering if anyone has considered what's going to happen when people get out of the habit of using stack overflow, aka, training the model. Sooner or later the labour put into training will have to be encouraged.