Artificial General Intelligence: An Upper Limit

The implicit or explicit goal of many areas of artificial intelligence research, since the dawn of the discipline, has been for machines to achieve artificial general intelligence (AGI) or strong artificial intelligence—that is, the ability to build devices capable of thinking like humans. Over time, tests have been devised, such as the Turing test, but today they seem largely obsolete, without this having implied the achievement of the goal.

It is difficult to say whether this dream will ever be achieved. The largely inadequate criteria for identifying the end of the process make it necessary, in my opinion, to define a limit that allows us to measure our understanding and our proximity to the goal. What follows is a personal and intuitive consideration, not a formal one.

I borrow a quote from Polish mathematician Marc Kac to illustrate a possible “upper limit” to which machines equipped with artificial general intelligence should aim. Kac talks about the physicist Richard Feynman (Nobel Prize winner for Physics in 1965) in his book “The Enigmas of Chance”, also reported in the biographical note at the beginning of the first volume of Feynman’s “Lectures on Physics”. If one day we can talk about a machine this way, then we’ll be close to the first AGI:

There are two kinds of geniuses: “normal” ones and “magicians”. A normal genius is someone like you, and I could be if only we were much better. There’s no mystery in how their brain works. Once we understand what they’ve done, we’re sure we could do it too. With magicians, it’s different. They are, to use a mathematical expression, the orthogonal complement of us, and the intentions and purposes of their brains remain utterly mysterious to us. Even after we understand what they’ve done, the process by which they did it remains completely obscure. They rarely, if ever, have students, as they can’t be emulated, and it must be terribly frustrating for a brilliant young mind to try to fathom the mysterious ways in which a magician’s brain operates. Richard Feynman is a magician of the highest order.