"AI safety" groups are part of what's described here: you might assume from the general "safety" label that organizations like PauseAI or ControlAI would focus things like data center pollution, the generation of sexual abuse material, causing mental harm, or many other things we can already observe.
But they don't. Instead, "AI safety" organizations all appear to exclusively warn of unstoppable, apocalyptic, and unprovable harms that seem tuned exclusively to instill fear.
But they don't. Instead, "AI safety" organizations all appear to exclusively warn of unstoppable, apocalyptic, and unprovable harms that seem tuned exclusively to instill fear.