Elon Musk on Reasoning by Analogy (this is often used with respect to the singularity)

How often do you hear that Ray Kurzweil’s predictions or singularity predictions are bullshit because some other predictions in the past were bullshit? I see this argument more often then any other, although its not really an argument but more of a quick way of dismissing the idea. The second most common argument is that people who think there will be a technological singularity are believers in a new “religion.” Both the argument from failed predictions in the past and the argument from “singularity is a religion” are reasoning by analogy.

I was confused when I first read that Elon Musk had argued against reasoning by analogy and for the record I do not think he ever said this in the context of the singularity and I do not know whether he thinks the singularity will happen. I was surprised by Musk’s criticism of reasoning by analogy because it is so common and so pervasive that I almost thought that there was no way to reason without reasoning by analogy. I realize now that reasoning by math or as Musk says by “first principles” is at least one other way to reason, it is perhaps the opposite of reasoning by analogy, and in my opinion superior.

No doubt a huge variety of people have been making predictions for many years with little success. Speculating on the future of technology is interesting. I submit that it is unlikely that very many people who made predictions in the past had a rigorous methodology although they no doubt had their rationalizations. Yet, people dismiss the predictions of someone like Kurzweil or Moravec because of the failed predictions of other people which had nothing to do with their predictions. It would make more sense to hold people accountable for their own track record and Kurzweil’s is not bad at all. However, the larger point is that dismissing Kurzweil’s prediction this way is reasoning based on analogy and by way of contrast Kurzweil’s argument is from first principles. Since, Gordon Moore’s paper, and perhaps the observation has been made sooner but at least as early as Gordon Moore’s paper we have had an observed trend in the semi-conductor industry seen in exponential growth. Kurzweil extended this trend through 5 paradigms of growth in computing. Kurzweil, further notes that an exponential trend can either continue or die out and notes the actual physical constraints that will actually bring such a trend to an end and notes that there is a lot of room at the bottom. This leads to the conclusion that a singularity will occur. Kurzweil’s argument for the singularity is entirely based on first principles reasoning, in the same way Elon Musk in the video reasons by first principles in the battery example.

Another type of person having never heard of the singularity and wanting to quickly see if it can be easily dismissed and thus thinking about it be avoided hears its promising of eternal life and all sorts of fantastic things and instantly dismisses it as a religion. We don’t have time to evaluate all the crazy things people say so its actually understandable that people avoid the question when they can. The simple fact of the matter is that reasoning by analogy is weak and will often lead to the wrong answer. Many things are similar in one way but not others. Kurzweil predicts greatly extended life (“eternal life” is a bit of a straw man) but for very different reasons than religions do. When things are close analogies are even worse. For example take the obviously wrong statement “a dog named rover is a cat.” Rover has fur, is a mammal, has dna and lives with humans. Cats have fur, are mammals, have DNA and live with humans. Therefore Rover is a cat. Reasoning by first principles of what a dog is would not lead to this mistake but its easy to make a bad analogy that would make this mistake if you don’t have a clear picture of what a dog is and cannot derive what a dog is from first principles. This example may sound contrived but children actually make this mistake often.

Reading this over I realize I’ve used a few analogies in just this short blog post. Analogies neither can nor should be completely avoided. However, I feel Musk’s simple point has opened my eyes and everywhere I look I see “bad analogies.” Therefore, Musk’s reasoning by first principles has been one of the most insightful and helpful things I’ve learned in recent memory.