I (and the MIT study) are not talking about severity in and of itself but rather likelihood (and, to a lesser extent, timing).as the source of concern re: risk.
The businesspeople funding the development of “AI” have demonstrated themselves fundamentally uninterested in – if not outright happy to promote – outcomes like racial bias in algorithms or privacy violations. To that end, they’re glad for us to worry about gedankenexperiments like Roko’s Basilisk (I’m done for now, aren’t I?) as a distraction.
