Feynman Lectures Chapter 4.2: Add or remove weights in a non-ideal machine? Excerpt:

A very simple weight-lifting is shown in Fig. 4-1. This machine lifts three units "strong". We place three units on one balance pan, and one unit on the other. However, in order to get it actually to work, we must lift a little weight off the left pan. On the other hand, we could lift a one-unit weight by lowering the three-unit weight, if we cheat a little by lifting a little weight off the other pan. Of course, we realise that with any actual lifting machine, we must add a little extra to get it to run. This we disregard, temporarily. Ideal machines, although they do not exist, do not require anything extra.

Figure 4.1:

This may be already answered by @mmesser314
here, but I want ask the question from a different point of view.
During the first part of the paragraph, we are told that "little weights" must be "lifted off" the left and the right pans to get the machine to "actually
work". Assuming that in this context lift off = remove, that makes sense. But then we are abruptly told that we should
add a "little extra" if we want to use an "actual" lifting machine. I assume that actual = in real-life and extra = additional weight or force.
So which one is it if the machine is not ideal - do we add or remove weights for it to work? Do we do both? First "lift off" a little weight, let the lifting and the lowering occur and add a "little extra" to do the reverse operation?
 A: First just a comment on the wording: Feynman says you need to remove "a little weight," not "little weights." I picture this as using a knife to shave a small amount of mass from one of the cubes. Of course the details of how you remove the mass doesn't really matter, but I just want to make sure it's clear that Feynman is not saying to remove an entire box from either scale.
A second overall comment is that I think you are reading Feynman at a very high level of precision, whereas Feynman tends to use very physical and "natural language" arguments. Just as a suggestion, you might find a different book (Landau and Lifschitz? Kleppner?) that gives more mathematical details to be less frustrating. Of course Feynman has a lot of insight and is worth reading, but I'm just raising this since this is one common reason people find Feynman to be difficult to follow (until you already have some background in what he is talking about).
Anyway here is how I interpret Feynman's words:

*

*"lift off": remove some mass from the balance.

*"actual lifting machine": one whose pivot has frictional forces that act on the balance, which dissipate energy.

*"add a little extra": add some extra energy (not mass).

Putting this all together, if you "lift off" (remove) a little (infinitesimal) mass from the left plate in your figure, then the balance will start to lift the three masses. However, if there is friction in the pivot of the balance that can dissipate heat (which there will certainly be if this is an "actual lifting machine"), then the balance will slow to a stop before the masses reach their maximum possible height. Therefore one must "add a little extra" energy (beyond simply removing an infinitesimal amount of  mass) to overcome friction and raise the three masses.
