He stresses that the human mind does not qualify as a completely ‘General Intelligence’ but lies somewhere on the spectrum between AGI on one end and ‘Narrow AI’ on the other. This is one of several reasons why he does not expect AGI to be achieved by mimicking the workings of the human brain.

He describes how our brains fool us into believing that we understand our actions and decisions when we don’t. And why modeling an AI too closely on the human brain might make it too, vulnerable to false notions.
He also says, ‘I think virtual worlds are going to be absolutely critical to the development of Artificial General Intelligence.’ As well as ‘Right now connecting AI’s to virtual worlds is probably the best way to get an AI to have a general human-like embodied experience.’