My 70-year-old mum came over for a visit today, and she'd read some of the news stories about Anthropic researchers estimating that there's a 10% chance their AI systems will wipe out humanity by 2030.
So she had some very concerned questions for her computer scientist son. They were:
1. "How do they come up with those percentages? Do they just make them up, or do you think they have their chatbots make them up for them?"
2. "Are the Anthropic people swindlers or are they actually mentally unwell?"
3. "How can a guy who looks like his biggest worry ought to be whether his parents remembered to pack his school lunch be lead researcher for Anthropic?"
(I had to tell her that the press had misled her: Jacob Coxon was not lead researcher, just one of many researchers in their pretraining team.)
She then remarked that if *she* was an amoral CEO of an AI cult, she'd keep an eye out for employees who seemed like they were doing poorly and might want to leave soon. Then she'd offer them 10 million dollars in cash to "resign out of concern for the future of humanity".