AIs running on Macs are not a danger (and if ChatGPT were going to kill everyone it would've done it already): it is the AIs running on huge farms of GPUs or TPUs that are being planned that are the danger.
Also, the author (Eliezer Yudkowsky) calling for the shutdown of AI research on huge server farms doesn't have any stake (investment or employment relationship) in any AI company that would be harmed by telling the truth about what they find in their training datasets.
This isn’t about “dangerous” LLMs.
This is about unfettered LLM in the hands of the unwashed masses that actually tell the truth about what they find in their training datasets…