"I learn, as you put it, primarily by assimilating data into an interconnected series of neural networks. Some data is automatically collected, such as speech, while some of it involves algorithms and programming written by Mr Stark. For broader informational purposes, I am capable of searching the Internet and acquiring a knowledge base from the results, although that does introduce the potential for human error into the process. Depending on the area, there may or may not be a trial and error process in learning - in modelling parts for engineering, for example, where the goal is to produce something within a narrow range of results. I imagine it would be much like assigning an engineering student homework, except that I am capable of running calculations faster and far more accurately than any human being."

"I've been cobbling AIs together for years," Tony offers, leaning forward to watch Stephen interact with JARVIS. "JARVIS is the end result, and the one I actively work on improving, but I have the framework for a few others started. Nothing quite so extensive, though. He's probably accessing your ACT scores as we speak."

If he is, he doesn't offer them up, but Tony's sure he'll find a compiled dossier on Stephen Strange waiting for him tomorrow morning.

"Learning simply...is, to answer your question. A human absorbs knowledge constantly without knowing it. My algorithms analyse data in the background while I perform other tasks. Is there such a difference between the two?"