How to understand Feynman's reasoning about perpetual motion?

He's basically saying assume you have some complicated system of weights connect by pulleys, and each weight can be in only one of two states: up or down. But you can trade off which ones are up and down, for example you can make 3 light weights go up by having one heavy one go down, and there are many other moves like this you can do.

Now his point is that you can't have a sequence of moves which takes you to a final configuration which is just the same as the initial configuration except one of the initial down weights is now up.

His reason why is that if you could do this, then you could use the energy from lowering the weight to generate electricity or whatever. Note that after you have lowered the weight, you are back in the initial state. So in particular, you could use part of the energy gained from the process to power a machine that moves the weights again. This would give you an unlimited supply of energy.

Your confusion was thinking that feynman was saying it was illegal to have a lifted weight in the final state. This is not what feynman is saying. He is saying that the final configuration cannot have extra weighs lifted. So if an initially down weight is up in the final state, there must be another weight which has been lowered from the initial state.

My interpretation:

The definition of a perpetual motion machine is one from which more energy is produced than consumed i.e. getting something from nothing.

Classic examples of perpetual motion machines are those which involve some sort of repeating cycle, and there is an expectation of excess energy in some form at the conclusion of each cycle. Among the examples are systems which lift and lower weights; at the end of a repeating cycle, there should be an excess of weight lifted i.e. potential energy available for extraction. This is impossible, because in its simplest form would be equivalent to a ball spontaneously rolling itself up a hill.