michaelcohen comments on The Main Sources of AI Risk?

[michaelcohen]

Maybe something along the lines of “Inability to specify any ‘real-world’ goal for an artificial agent”?