From the original paper:

> However, it must be emphasised that this does not include other dangers posed through the misuse of these models, such as the use of LLMs to generate fake news. Similarly, we do not contend that future AI systems could never pose an existential threat. Instead, we clarify that, contrary to prevailing narratives, the evidence from LLM abilities does not support this concern.

I find myself unconvinced that LLMs are "inherently controllable, predictable and safe."

What are the odds this "study" was produced by an AI? ;)

> [LLMs] cannot learn independently or acquire new skills, meaning they pose no existential threat to humanity.

Interesting… guess I really don’t need to worry about meteors, greenhouse gasses, zoonotic diseases, nuclear weapons, or, you know, time and entropy.