Robots take in observations. They make theories that explain their observations. Different robots will make different observations and communicate them to each other. Thus, they will talk about observations.

After making enough observations they make theories of physics. (They had to talk about observations before they made low-level physics theories, though; after all, they came to theorize about physics through their observations). They also make bridge laws explaining how their observations are related to physics. But, they have uncertainty about these bridge laws for a significant time period.

The robots theorize that humans are similar to them, based on the fact that they have functionally similar cognitive architecture; thus, they theorize that humans have observations as well. (The bridge laws they posit are symmetric that way, rather than being silicon-chauvinist)