Sorry Sydney: Generative AI Doesn’t (and Shouldn’t) Have a Liability Shield

This is such a disingenuous argument. Trying to frame it as if LLMs were a toxic mine, where the company profits while contaminating the water supply.

What the author is really doing is treating people like children, as if they needed to be protected from the possibly mean and false statements of the LLM, even after they are warned that it can be mean and dishonest. That unless the companies can guarantee it never lies or returns private information (that was on the open internet..) or produces bias or discrimination (what a long and vague list of harms), they should just not deploy it.

Basically, we shouldn't have LLMs until they can be completely baby-proofed. I.e. never.