Will Human-level AI Emerge from Virtual Worlds?
I attended Collision’19 in Toronto between May 20–23 recently. One of the most impressive speakers was Stephen Wolfram of Mathematica and Wolfram Research fame.
For those who have studied science or engineering, Mathematica is probably a familiar software tool, but for those not familiar, that’s just say that if Sir Isaac Newton was the man who organized Calculus into a book, Stephen Wolfram was the one who made a calculator for Calculus.
Besides the very foundational contribution of making mathematics accessible and easy for people to compute. Stephen also designed the Wolfram Language, which is a way to perform computations using natural language and existing human knowledge.
According to Stephen, before math notations were invented, it was really difficult to learn or even discuss about math, but once people have a common abstractions on which to capture and discuss the knowledge, progress becomes greater. Wolfram language aims to do the same for computation.
During the presentation, Stephen announced that the engine is now freely available to the public, and you can even try it online!
Although I knew about Stephen and Mathematica, they are not part of my daily life, and there was only a distant respect for the man. But that day, after watching him showing the audience first hand what Wolfram Language could do, I was deeply impressed and had great respect for the man.
After his talk I went to ask a question that had been on my mind: “will AI advance faster within virtual worlds than in the real world? As all the representations are already fairly digitized and fully accessible to the machine?”
His response? He agreed that virtual lands are more readily accessible, and that helps somewhat with the knowledge representation efforts, at least for example, you don’t have to pattern recognize whether that shape is a person or a cat, in order to perform higher-level cognitions.
However, it’ll still be a challenge, to recognize that the three or four man standing and talking, whether in physical or virtual space, are actually “having an argument” or “having a conversation”. Or whether they’re “chatting” versus “teaching”.
That kind of higher-level abstraction will still be difficult to recognize and identify, yet would be important for knowledge representation and cognition.
Besides my question, he answered questions from a group of people surrounding him, and I was impressed to find that he could always provide an answer that was both sensible and responsive to the attendee’s line of thinking.
In a sense, he had an answer to any question, from the most trivial to the complicated, across different knowledge domains. That to me, is a sign of true intelligence.
It was a pleasure to hear the man speak, and truly an honor to be able to share in his wisdom! If you’re interested to learn more, I highly recommend a reading of his own blog on what Wolfram Language is, and why it is important.
