[AppleJus]

If you are to make a pure - in the ideal sense - AI agent, you will have to create the entire social life of that agent, to the least, and most probably if not surely the whole world. This is what it takes and is the only way to mimic exactly a human brain state at a given time. The qualia won't be accessible otherwise. And even then you won't be able to because of the causality principle (or butterfly effect as some people like to say - not always appropriately), so you are essentially always left behind trying to catchup (and the use of the supercomputer speed won't be relevant, since the AI agent always - and minimally - depends on a physical environment.)

What your thoughts on this?

--Edit: The more interesting question would be where to draw the line? Where to stop in the experiment / true-AI description I made above?
In other words, what is the strongest yet realistic AI we can achieve - between true ideal AI and weak rudimentary AI? Deep learning seems to be an answer, but i believe it is not, mainly because it produces some results irrelvant to human intelligence.