Brilliant People, Brilliant Machines: Introduction
You are reading the first installment of a blog series that goes through five steps that brilliant people use to be brilliant, in an attempt to make useful connections with actual and potential methods of improving Artificial Intelligence(AI). In each installment, I break down one thing that brilliant people do and attempt to connect it with actual, or potential AI strategies and methods. Each of these steps is meant to deepen our current understanding of what AI can do. The main goal of these series is to spur novel thought in our field and think about AI differently.
In conversations about AI (artificial intelligence) there is often an implicit assumption that a bigger faster computer is a smarter AI. And that all we really need to do is to make a big-enough fast-enough assemblage of transistors and feed it enough data, no matter how inconsequential and voila, a conscious sentient being! Then, we just make it a bit faster and a bit larger and BAM — we have a new god among us and we become mere ants without moral weight.
In the understated language of a scientist, the above scenario is unlikely. First, at both the microscopic and macroscopic level, neural nets are to brains what a sphere-in-vacuum is to a cow. Now, that’s not useless. Physicists and engineers solve an insanely large number of problems in the universe by making that assumption. E.g., A sphere-in-vacuum assumption will do just fine if you want to aim a cow (or a human) over the wall with a trebuchet. But it’s a ways away from making a brand new herbivore. While turning up the size of a neural net to 11 will lead eventually to some form of emergent phenomena, there is clearly a vast space of possibility in which that emergent phenomena is not recognizably sentient.
While turning up the size of a neural net to 11 will lead eventually to some form of emergent phenomena, there is clearly a vast space of possibility in which that emergent phenomena is not recognizably sentient.
Further, beyond the structure and physical/chemical method of the brain’s function, what the brain is actually attempting to do is (believe it or not) quite important. While some attention has been paid in the literature to more closely mimicking what we know of the brain at the cellular level, seeking inspiration from the high level functions of the brain has been out of fashion since the direct importation of logic failed to produce a mechanical logistician.
Logic and reasoning is at the base of our conscious mental toolset. In this series I’ll be going straight to the top of the tower and trying to look at the high level tools of the brilliant through the eyes of a modern AI approach.
I’ve been breathing some pretty rarefied air over that past decade and a half. We’re talking about an assortment of top Physicists, Mathematicians, Computer Scientists, and Nobel Prize Winners. When you’ve worked with as many brilliant people as I have, you notice some patterns. There is a lot of information out there on productivity. But I think the discussion on brilliance is too often ignored, probably because people have the false idea that brilliance is a born-in trait. In this series of posts, I’ll walk you through the key steps in the learning process of the brilliant. For each step we will also discuss what analogous processes exists in the world of AI and Machine Learning. Hopefully you will take something way from both sides and improve your own thought process, as well as your AI’s
Hopefully you will take something way from both sides and improve your own thought process, as well as your AI’s
Later in in the series we’ll come to some “secrets” of the brilliant that I have not seen discussed or taught, yet all the best minds I know use them. Interestingly, among the “secrets” for human brilliance are some of the most routine machine learning tasks.
For this first part we will be looking at something closer to “common sense”. It is a truism that extraordinary results require doing things out of the ordinary. Often overlooked are ordinary things that are not ordinarily done. And among the “common sense” human techniques are some of the most far out visions for artificial intelligence. Whether it is a general truth or not, I can’t say. However I can say that it pays to pay deep attention to those instances where humans and machines have vastly different competencies. Without further adieu, here is part 1 on the base layer for brilliance: the table stakes required for normal (but not common) human functioning:
Part 1 — Table Stakes for Humans (Small step for Humans, Giant Leap for Machines)
