AI x-risk reduction: why I chose academia over industry — LessWrong