

Every “AI” company should be (reasonably) liable for what their tools say. How is this even a novel idea?
If a newspaper accidentaly prints false information, they have to publish a correction and might pay a fine. But if they print a front page article about making a pipe bomb, then the editor would probably get sentenced.
This approach would be perfectly fine with LLMs. I understand the nature of the technology, but if they cannot guarantee the quality of the output, then the product is just not ready yet, and we are currently doing unpaid testing.







More like vaping, as it removes some of the harm (adslop and horrible UI), but keeps other parts that are still bad for you (current reddit itself)