[bookaway]

Seriously. Considering the existential risk the AI Safety tribe says exists if sufficient care is not taken, they appear to be existentially incompetent in their messaging. It should be their number one priority to convince the legion of competent non-AI engineers who will be developing and running the infra for their systems to come to their side. Currently they seem to be failing spectacularly.

If it wasn't for others in the AI community vouching for their chops it'd be easy to mistake them for Philosophy of Mind majors going by their tweets. They need to convince their fellow engineers not politicians. Right now when people hear "safety" for AI they think "regulation by incompetent bureaucrats".

The AI safety people need to remember that "you can't tell people anything", you have to show them [0].

[0]http://habitatchronicles.com/2004/04/you-cant-tell-people-an...