In the Moon Versus Mars! thread, an objection to sending people to a permanently manned Moon Base is that robots can do practically anything a human can do, and cost less, because they somehow don't require an infrastructure, whereas humans do. Naturally, my own position is that there is no way robots by themselves--or even teleoperated robots by themselves--can accomplish anything beyond a mere proof-of-concept pilot project demonstrating the feasibility of some industrial process or other on a place like the Moon.
I am using the term "robot" in its loosest sense as denoting any powered machine capable of replacing human labor. In that sense, a bulldozer is a teleoperated robot where the operator just so happens to be riding on top of the robot.
As a concrete example, consider an industrial plant on the Moon capable of generating several hundred tons of LO2 per year to be used for refueling orbital fuel depots. Such a plant would require that several modules sent from Earth would have to be bolted together. Do you think that robots alone could bolt together the required modules, maintain the plant, and then deliver the produced LO2 to orbital fuel depots using SSTO's that are themselves maintained by robots?
That is the poll question, but the discussion need not be limited to LO2 plants. Can robots do lunar geology and prospecting as well as humans? If not, are humans worth the extra cost? Perhaps there are important things that need to get done that can't be accomplished by robots at any price. In that case, humans are infinitely cheaper.
But let's stick to the realm of what is possible within the next few decades. Historical evidence, analysis of working examples of industrial processes on Earth, and mathematical engineering data is preferred to evidence taken from pulp science fiction.