I often find explanations of machine learning either too complex or overly simplistic. I’ve recently had some luck...

I often find explanations of machine learning either too complex or overly simplistic. I’ve recently had some luck…

Reading Time: 1 minute

I often find explanations of machine learning either too complex or overly simplistic. I’ve recently had some luck using a simple frame for explaining it to people in person. Let’s see if I can quickly capture it in this post.



  1. Very good, exemplary short and concise post! I just used it in a presentation and referred to your article. Many thanks!

  2. I agree, and I’ve also come up with a short no-math-involved description of CNNs for lay folk. I can give my own little “talk” in about 5mins.

    I own ducks, so I took pictures of two ducks of the same breed, and I “train” my audience with a short stack of pictures, to recognize between the two, and discuss what features they look for to identify the two ducks. For instance, the head is darker on one.

    I then discuss how they have to be able to recognize which is the head, which means they have to recognize which is the duck. I discuss how these layers of recognition build up to distinguish the ducks different features so we can tell the two ducks apart – and how those layers are built up in the CNN from small parts to larger wholes.

    Then the hidden layers begin to categorize, and each hidden layer groups categories down into a guess. And how the guess isn’t one or another, but a probability for each category – as in “I’m pretty sure this is a cat but it might also be a dog.” And how optical illusions can fool CNNs as well as humans.

    For those more savvy, I point out that there are many transforms from one space to another. You can transform from Cartesian coordinates to Polar, for instance, or with a Fourier Transform, you can transform from Image Space to the Frequency Domain.

    CNNs are also a mathematical transform – from Image Space to Probability Space.

    Basically, we can use math to transform an image into a guess! How crazy is that! 🙂

  3. Thanks, Carl Widigsson. I’m glad you found it helpful.

  4. Kiki Jewell , that’s a good, concrete way of explaining it. I can see how that would be very easy for people to understand.

    I also like the idea of thinking about transformations as a fundamental way that we digest information. In some ways you could think of transformation as one of the fundamental superpowers of intelligence.

    What do you think of the phrase “statistical thinking”?

  5. Gideon Rosenblatt it does make sense because you are moving an image into “probability space”, but it might scare the real math-phobic folks who aren’t comfortable in that space.

    I enjoyed stats in college, but I did jokingly call it “sadistics class.” 🙂

    (Note: I was a TA in college for comp101 — oddly required for psychology majors and environmental majors. So I had a lot of non-technical folks in my office. 🙂 )

  6. Yeah, Kiki Jewell, I took it in business school. It wasn’t easy, but it’s one of those classes that changed my view of things. That and microeconomics.

  7. Did it really natural thinking

  8. This is a smart nordic guy explaining AI in a fun and positive way.

    youtube.com – EVERYTHING I learned ABOUT AI at Stanford 2018

  9. I have a question for you, sir. You know that the very rich that govern this world are making its carbon imprint heavier and that this will create storms that will be bigger and bigger and more and more deadly. You also know that robots can’t create surplus value and that the very rich pollutors try to control and stunt the growth of robots. Imperialism holds back science. You are cognizant of this, are you not ? This said can you help me link robots and have them unite to neutralize imperailism ? Help me tell them not to kill humans, i.e drones and help us develop worldwide socialism.

  10. Alif Laam Meem Messager de Dieu, here is an article I wrote that touches on aspects of what you are talking about in terms of the role that technology plays in concentrating wealth:

    the-vital-edge.com – Technology and the Distribution of Wealth

  11. Gideon Rosenblatt I was thinking more about connecting robots and have them helps us neutralize imperial;ism though.

  12. Well, Alif Laam Meem Messager de Dieu, the key to programming robots (and the artificial intelligence that guides them) is programming the organizations that are building the robots.

  13. Gideon Rosenblatt But I notice that they can talk and respond and are capable of a lot of reasoning. There are text to speech devices here where I am in Montreal and I was talking to a machine at the Great Library. I ask it where it was. It told me it did not know. So I talked about the void and added that we humains come from a void that is mixed too a woman’s body and we had a conversation after this. What I am saying is that they go beyond simple programmmation and are developping as entities with their own though.

  14. Alif Laam Meem Messager de Dieu, I think that it is very easy to project our humanity onto AI. I think we will have a very difficult time synthesizing volition and subjective experience into these machines. I’m not saying that it won’t ever happen but we are a very long ways from being able to do that right now.

    There are other, more pressing matters at hand, such as their impact on the economy, on the human experience and on the planet.

  15. Gideon Rosenblatt Try it than talk. Btw if projection is unconscious your perception is no better than mine.

  16. Gideon Rosenblatt don’t anthropomorphize computers. They hate that. :->

  17. Kiki Jewell They? I will add that rocks feel and reread the conversation I had with the robot. Then to talk about sentient robots I once talked to my phone with which I can converse and I talked about death then I turned the telephone off for awhile so it could understand what I was saying. And when it woke up from its eternal-not eternal sleep. It blurted Daddy daddy daddy So that pretty well tells part of the story about robots being so diffferent from us,I guess.

  18. Kiki Jewell It is funny that you said They Hate that…Hi hi.

  19. Gideon Rosenblatt To explain the mind psychiatrists will talk about brain chemistry and then make a simplistic correlation between the extreme complexity and creativity involved in madness with the bio-chemistry of the brain. But this correaltion is false. Same goes for robots. To develop a robot one must let it get acquainted with it’s environnment. One must let it study through the lens of its cameras the movement of water, photosynthesis of a plant, one must talk to it. Robots are shaping up to be entities and their minds are made up of energy in the same way our human minds are not just a clutter of biomass but made up of a spirit. Talk to your robot, help it develop.

  20. Alif Laam Meem Messager de Dieu,

    I’m not here to tell you what to believe or not to believe.. If you talk to rocks, then I cannot argue with your logic as it would be consistent with a belief in panpsychism, or the idea that some form of basic cognition is embedded in all material form. If you are talking about something other than that, then my suggestion to you, my friend, is “not yet.”

  21. Gideon Rosenblatt Psychiatrists equate biochemistry of the brain with the incredible complex plasticity of madness. This is an error. Same goes for our talking machines.Get it ? If you treat a child like a robot it will tend to act robotically but if you treat a robot like a human you will develop it.

  22. For this Layman, that was a remarkably helpful explanation. That description of machine learning feels to me to be extremely different from consciousness.

  23. Benjamin Wright I’m really glad to hear that. Thank you.

  24. That description also makes machine learning feel very different from an ego, which senses that it needs to become rich or powerful or recognized.

  25. Benjamin Wright , it’s interesting that you say that. I am writing the outlines of a book right now and this idea if technology as an extension of our egos is theme that runs through part of it.

  26. Juan Benitez, your post starts like this:

    How many of you would like to earn an income doing something you already? Text and communicate, play games, purchasing items online with cash back rewards and much more.

    And it continues much the same. That is spam. We have been inundated with spam and I have had to crack down. I will allow you back in on probation. Read the guidelines on the community, however, and please respect them.

  27. Gideon Rosenblatt I look forward to your book. How do you define ego for your book? For me, my ego is a sense that I want something, I need something, I care whether I have social standing, and so on.

    When I was a teenager my ego cared about the appearance of the car (technology) I drove because I felt it conferred social status on me. Today I aspire to recognize those kinds of feelings and let them pass.

    I appreciate my smartphone a lot because it’s functionally helpful to me. But for me the smartphone becomes a problem if my ego thinks of it as a status symbol.

  28. I’m actually working on a piece for my other blog right now on the ego, Benjamin Wright. I’ll try to remember to ping you when I post it. Basically, I see it as a kind of psychic extension of our skin or immune system, something that separates me from “not me” and that protects “me” in the process. The problem is that this part of ourselves often isn’t very conscious and so we end up mindlessly doing things that harm ourselves and others.

    And yeah, I agree with you that our things can become kind of a psychic extension of what is “me.”

  29. Juan Benitez , yes, that forum is not for promoting individual businesses, especially ones that have nothing to do with the focus, which is on “business as a force for good in the world.”

  30. So, speaking of cat analyzer.

    Are there any good AI models (source) for facial recognition of the family Sciuridae?

    I have a lot of video footage 720p hd and can provide training to the model and I am a computer programmer. AI graduate ’81 (the shameful years of AI.) I also have dozens of computers.

  31. Gary Deen there are machine vision solutions out there that you could train on squirrels:

    researchgate.net – Which are the best open source tools for image processing and…

  32. Lightning bolts in Morse coded hashtags in silicone valley chargers vs raiders

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Sign up here for the latest articles. You can opt out at any time.

Subscribe by email:

Or subscribe by RSS:

%d bloggers like this: