523. Inner Goodness – 529. Efficient Cross-Domain Optimization

523. Inner Goodness

If you force people to be moral, they are less likely to actually be or act moral. Morality shouldn’t be a duty. It should come from within. You should want to help other people.

OTOH people who are moral, although they actually hate being moral, are more admirable than people who enjoy being moral. But the point is not to be admirable. The point is to be moral.

524. Meetup Post

525. Expected Creative Surprises

If you deal with beings that are more intelligent than you are, you won’t know which specific actions they will take. If you knew that you would be at least as intelligent as they are. But if you know their goals, you’ll expect a specific outcome, namely the realization of these goals.

E.g. if you play chess against Kasparov you won’t know which moves he will make, but you know that you’ll lose.

526. Belief in Intelligence

Similar to the last post.

527. Aiming at the Target

Intelligent beings, or more general, sufficiently powerful optimization processes try to steer reality into a specific region of possible futures. Often a very narrow region.

528. Measuring Optimization Power

Discusses how to measure the power of an optimization process.

529. Efficient Cross-Domain Optimization

Yudkowsky defines intelligence as efficient cross-domain optimization.

In other words:

So again I claim that this – computationally-frugal cross-domain future-steering – is the necessary and sufficient meaning that the wise should attach to the word, “intelligence”.


Leave a Reply