We should have seen AI coming. We know from multiple regression that correlations can be used to predict anything from how much you weigh to how much it will snow in December. If correlations can predict that, they can predict anything, it’s just a matter of getting the data and intelligence could arguably be defined as the ability to predict.

However just because AI can pass the bar exam, hit the ceiling on the verbal sections of the WAIS-IV, write plays in the style of Shakespeare, compose music that resembles Mozart and create drawings in the style of Picasso, does not necessarily mean it’s more intelligent than the average man. It achieves these great feats by processing the entire corpus of civilization. If it looks taller than us, it’s because it’s standing on the shoulders of our giants.

The only way to know if AI is truly more intelligent than people (or even as intelligent) is to personify it. Put it in the cranium of a humanoid robot that is indistinguishable from a human in the way it looks, moves, feels and sounds. Have it perceive the World through artificial eyes, ears, and touch. And then give it just one goal: get as much money as possible.

At first people would assume the AI was a mental retardate because it would behave awkwardly in its humanoid body, but it would quickly learn that acting human was financially incentivized. Would this AA eventually become the World’s first trillionaire (IQ 150+) or would it just languish in jail because it was caught stealing (IQ 90)? Of course luck is a huge factor in wealth so we’d might want to run the experiment on many such robots and observe the median outcome.

Another thought experiment would be if we get all humans (or at least all the smart ones) to leave the Earth and replace them with millions of AI humanoid robots that had only one goal: Get to to the moon. I agree with Stephen Jay Gould and RIchard Klein that the modern human mind (average IQ around 90) more or less evolved about 40,000 years ago (the age of the first true cave art), so it took humans about 40,000 years ago to go from the stone age to the space age (and most of us weren’t even trying).

How many years would it take millions of humanoid robots to achieve the same shared goal. And how would he program even “incentive” them to do so? Perhaps they’d get a “pass” for every step closer to the moon like jumping in the air or climbing a mountain, and get a fail for every step further from the moon like falling in a hole.

Given 40,000 years, would their shared neural network modify its algorithm to the point where it reached the moon? If they did it in half the time humans took, we could argue they’re roughly twice as smart as humans (IQ 180, and yes IQ can be multiplied) or would it take them over 80,000 years, implying an IQ below 45? I realize these are not perfect apples to apples comparisons since unlike AI, humans had other problems to solve like getting food and reproducing and the subset of humans who reached the moon evolved IQs somewhat above the 90s, but these points roughly cancel each other out.

The above art was created by AI