What I got from fiction is an early notion of an efficient goal-reaching agent.

Two particularly good examples that influenced me the most are Gully Foyle of "The Stars My Destination" ("Tiger! Tiger!" in British edition, my favorite fiction work), and T1000 in Terminator 2. To summarize the insights I got from these works:

• Agents differ in their goal-reaching ability – some are better, some are worse.
• It is certainly desirable to improve my own goal-reaching ability.

It may be that the first insight, combined with non-fictional insights I gained from "The Selfish Gene" and my own programming / software development experience, later helped me comprehend the notion of a non-anthropomorphic optimization process.

(T1000 may seem a bad example of an efficient goal-reaching agent because it failed to achieve its goal – but let's remember than it's fiction. A real-world Terminator 2 scenario would have ended the very second T1000 got a clear line-of-sight to John Connor – let alone the fact that Skynet could simply send a ticking 100-megaton nuke instead of a killer robot, or, if it wanted surgical precision, terminate Connor while he was still a baby.)