Human beings, courtesy of the gift of cognition, are capable of creating reusable data, information, knowledge from simple or complex observations in an abstract realm. A machine on the other hand can only discover and infere based on a substrate of structured and interlinked data, information, or knowledge in a concrete human created realm e.g., a Web of Linked Data.

As is quite common these days, Yihong Ding has written another great piece titled: A New Take on Internet-Based AI, that delves into this specific matter. Yihong expresses an vital insight as excerpted below:
"Artificial intelligence is supposed to let machines do things for people. The risk is that we may rely too much on them. Two months ago, for instance, writer Nicolas Carr asked whether Google is making us stupid. In my recent blog series "The Age of Google," I extended Carr’s discussion. Due to the success of Google, we are relying more on objective search than on active thinking to answer questions. In consequence, the more Google has advanced its service, the farther Google users have drifted from active thinking."
"But at least one form of human thinking cannot be replaced by machines. I am not talking about inference/discovery (which machines may be capable of doing) but about creation/generation-from-nothing (which I don’t believe machines may ever do)."

I tend to describe our ability to create/generate-from-nothing as "Zero-based Cognition", which is initially about "thought" and the eventually about "speed of thought dissemination" and "global thought meshing".

In a peculiar sense, Zero-based cognition is analogous to Zero-based budgeting from the accounting realm :-)