Pattern comments on Drexler on AI Risk

[Pattern]

Can you construct agents that are guaranteed to ‘achieve Pareto optimality’?