The 4 things it takes to be an expert

Published 2022-08-02
Which experts have real expertise? This video is sponsored by Brilliant. The first 200 people to sign up via get 20% off a yearly subscription.

Thanks to and Chessable for the clip of Magnus.

Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive psychology, 4(1), 55-81. –

Calderwood, R., Klein, G. A., & Crandall, B. W. (1988). Time pressure, skill, and move quality in chess. The American Journal of Psychology, 481-493. –

Hogarth, R. M., Lejarraga, T., & Soyer, E. (2015). The two settings of kind and wicked learning environments. Current Directions in Psychological Science, 24(5), 379-385. –

Ægisdóttir, S., White, M. J., Spengler, P. M., Maugherman, A. S., Anderson, L. A., Cook, R. S., ... & Rush, J. D. (2006). The meta-analysis of clinical judgment project: Fifty-six years of accumulated research on clinical versus statistical prediction. The Counseling Psychologist, 34(3), 341-382. –

Ericsson, K. A. (2015). Acquisition and maintenance of medical expertise: a perspective from the expert-performance approach with deliberate practice. Academic Medicine, 90(11), 1471-1486. –

Goldberg, S. B., Rousmaniere, T., Miller, S. D., Whipple, J., Nielsen, S. L., Hoyt, W. T., & Wampold, B. E. (2016). Do psychotherapists improve with time and experience? A longitudinal analysis of outcomes in a clinical setting. Journal of Counseling Psychology, 63(1), 1. –

Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363. –

Egan, D. E., & Schwartz, B. J. (1979). Chunking in recall of symbolic drawings. Memory & Cognition, 7(2), 149-158. –

Tetlock, P. E. (2017). Expert political judgment. In Expert Political Judgment. Princeton University Press. –

Melton, R. S. (1952). A comparison of clinical and actuarial methods of prediction with an assessment of the relative accuracy of different clinicians. Unpublished Ph.D. thesis, University of Minnesota.

Meehl, E. P. (1954). Clinical versus Statistical Prediction: A Theoretical Analysis and a Review of the Evidence. University of Minnesota Press. –

Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus and Giroux. –

Special thanks to Patreon supporters: RayJ Johnson, Brian Busbee, Jerome Barakos M.D., Amadeo Bee, Julian Lee, Inconcision, TTST, Balkrishna Heroor, Chris LaClair, Avi Yashchin, John H. Austin, Jr.,, Matthew Gonzalez, Eric Sexton, john kiehl, Diffbot, Gnare, Dave Kircher, Burt Humburg, Blake Byers, Dumky, Evgeny Skvortsov, Meekay, Bill Linder, Paul Peijzel, Josh Hibschman, Timothy O’Brien, Mac Malkawi, Michael Schneider, jim buckmaster, Juan Benet, Ruslan Khroma, Robert Blum, Richard Sundvall, Lee Redden, Vincent, Stephen Wilcox, Marinus Kuivenhoven, Michael Krugman, Cy 'kkm' K'Nelson, Sam Lutfi, Ron Neal

Written by Derek Muller and Petr Lebedev
Animation by Ivy Tello and Fabio Albertelli
Filmed by Derek Muller and Raquel Nuno
Additional video/photos supplied by Getty Images
Music from Epidemic Sound (
Produced by Derek Muller, Petr Lebedev, and Emily Zhang

All Comments (21)
  • BUDA
    The pattern recognition became very clear to me when I learned Morse code. The human brain takes 50 milliseconds to process and understand a sound. People regularly send and receive Morse code at 30 words per minute, which puts the dit character and the gap between all characters at 40 milliseconds. So you literally have to process sounds faster than the brain can recognize them. Over time you start to hear whole words in the code rather than individual letters, but you still have to decode call signs character by character. You basically cache the sounds in your brain without processing them, and once the whole set of characters passes, your brain is able to turn it into an idea and add it to the stack of previous ideas while your ears are already caching the next set of characters.
  • Lucas Carman
    Getting comfortable is the part that always kills me. I learn very quickly but once I get something down fairly well, I stop challenging myself and just rest on that success.
  • As a trained physicist this was really interesting. I have not the best memory recall, some guys know the answer to a problem they did years ago, but I always have a „gut feeling“ how the equations will emerge and I can see a strong pattern in equations, even looking at it for a small amount of time is often enough to restructure the stuff in my head - even when not perfect, it’s a good cope for a usually bad memory recall
  • CutiePi
    After having read Moonwalking with Einstein, Fooled by Randomness, Sapiens and Thinking fast and slow. This really felt like a condensed version of parts of each book combined.

    Very good video, cheers !
  • Jessica Tatum
    100% this is how I was trained to be a ballet dancer and I didn't even recognize it. We do the same movements in slightly varying patterns every single day in a structured class, and for actual repertoire we repeat the EXACT same movements over and over, with a teacher or coach telling you what to improve after each attempt. As you get stronger, you do more and more challenging combinations of movement with increased complexity and strength requirements, and you spend more time reviewing and conditioning on your own time. Eventually you get really good at learning and doing choreography in certain styles/from certain choreographers because you start to recognize the patterns of movement they tend to employ.
  • This video is thousand times more helpful than a thousand self-help motivational videos here on YouTube,,, thanks 🙏
    Wow, this was incredibly insightful!
  • Personally I learned this lesson with driving. I consider myself a great driver with 15 years on the road, but when i started racing 100mph karts 3 years ago i quickly realised that the 15 years of comfy driving was absolutely worthless in terms of racing near the limits. I am getting my ass kicked by teenagers who have never driven a car. After 3 years of karting myself i can proudly say that im still not even close to catching up to them. They have pushed the limits for years despite their age. In terms og driving, lets face it, they are the experts.
  • I couldn't justly articulate the relief emotionally, intellectually and spiritually this video has provided. Very happy you're continuing to follow the philosophical thread of chunking.
  • zyoface
    To steal from my high school teacher, "practice does not make perfect. Perfect practice makes perfect." This also seems to be an apt checklist for composing a well performing machine learning agent (or at least are 4 very relevant circumstances to consider). Interesting as always, thank you!
  • Peter Obara
    As a radiologist, the feedback is there but you have to put in a lot more effort and discipline to seek it out (save and follow-up on tough cases). Part of the deliberate practice becomes going back through case collections that have follow-up already so the feedback becomes more immediate. This way you can go through hundreds of cases pretty quickly (if you put in the effort), much like a chess player reviewing old games.
  • AlanKey86
    This is a very timely video for the start of a new college term in September - I'll definitely be showing this to my new students!
  • I've had years of formal classical music training, work in a competitive STEM field, basically a lot of experience with deliberate practice. It's funny though, the one skill that emphasized what this video was about more than anything else was butterfly knife flipping. The blade cuts you if you mess up, everything depends on how your fingers move and how you judge inertia, tricks are clearly defined and compartmentalizable, and if you don't learn new tricks, you are only stuck with the specific ones you know. It literally teaches you how to practice
  • rachelle2227
    I am a paper quilling artist that sells my work/done it semi professionally for a few years. It was interesting watching this, thinking about how this applies to art! I hadn’t really started experimenting and trying harder until I started doing it professionally (and before , a lot of the times I was a kid/teenager/busy college student). Showing my work publicly and looking at other artists, in my own niche and outside it, has really improved my designs/skill. Everything in this video was so true! What’s fun about being an artist is, I don’t think there really is a cieling, there’s always a new look/idea/ or technique you can explore with art.
  • ObviouslyASMR
    These four things actually made me think of what's required for AI to learn something

    1. Repeated Attempts with feedback:
    This represents the data you've got and its labels, or alternatively actual attempts and rewards in the case of reinforcement learning. Examples are AI "attempting" to classify images of dogs/cats and being told it's right or wrong. For RL it could be a chess playing AI playing a match and receiving feedback in the form of winning or losing.

    2. Valid Environment:
    Like illustrated for humans in the video, if the environment is random, AI won't end up learning anything. However in the button example, unlike the human, and much like the rat, the AI should converge to only choosing the green button. If the input to the AI is the same for each attempt this convergence should happen pretty much immediately. If the input is completely random the AI will still attempt to find patterns like the human, but I believe it should still converge to only the choice with the highest probability.

    3. Timely Feedback:
    This one isn't super relevant to supervised learning since it's assumed the data is provided with labels, no matter how timely those labels/feedback were given in the data gathering stage. For example, if you're using AI to predict grades of students based on their past grades, you won't wait for current students to get their grades; you'd simply train on past data that already includes the feedback (which I guess is also possible for humans actually).
    Timely feedback is more relevant to reinforcement learning. That's why RL training is often done in simulations instead of the real world: to speed up feedback (and also to limit actual real world damage in some cases, like in robotics or self-driving). An example of this sped up feedback is the famous AlphaStar playing a custom version of the StarCraft game, where instead of waiting for the standard game in real time, it was able to play something crazy like 400 years overnight.
    However, this is actually just learning faster. The AI would learn just as well if it didn't have timely feedback, just slower. For humans I guess the problem is that we might forget our thought process and decision by the time the feedback arrives, because our minds are occupied with other things and not focused on a single goal.

    4. Deliberate Practice / Don't get too comfortable:
    Indeed if AI comes across certain rare examples, and then keeps training afterwards without seeing similar examples again, it can end up "forgetting". This happens because the model is continuously optimized for the most recent batch of examples (in mini-batch gradient decent at least), which means at every nudge of its parameters it does not care at all about previous batches. The reason this works is because the batches have patterns in common, and therefore the model will learn these common patterns, and gradually forget anything that doesn't line up with these patterns. This is a feature, not a bug; if it learned everything, instead of just the patterns, that's simply memorization, which doesn't generalize to new examples not included in the data.

    Basically, going by this video, if a human can be an expert at something, AI can do it better (if the data is there) 😌
  • A J
    "Only after a refresher course could the (20-years experience) doctors accurately diagnose (rare) diseases" better than recent med-school graduates.
  • I used to play tournament chess in HS, learned tennis as an adult and played in USTA leagues, learned a few instruments and played in a few bands. The number one fail I saw of people along the same journeys as I while learning new things is their comfort level. Everyone has a rough time learning but some would gain a little competence and rest on those tiny laurels - and not get more competent. It seemed like people would find the laurels that fit their egos and then they stopped. They didn't go until they exhausted their abilities. Someone/something didn't say stop. They stopped themselves.
  • I recently had a MASSIVE argument with my university because they repeatedly did not provide any feedback to essays or exams. Just a mark and that's it. I backed my perspective with a ton of academic works on education, that I doubt any of them ever read.

    I'm going to show them this video. Because university courses that don't provide feedback are virtually useless.
  • Dangerous Dave
    Fantastic video! After 25 years of Air Traffic in the busiest facility in the US, I was able to recognize when someone would make it through a training program versus someone that would need to go to a lesser facility or even let go. All I needed was an one hour session. I had a 100% success rate that baffled my management and training program. It was simple to spot but difficult to explain. After watching someone on a training session, they would slowly transform their thinking exactly the way you described in the video. You could see growth and understanding and occasionally experimenting with techniques. The folks that didn't make, showed very little to no change. Every person in training would get 100's of hours of training, yet it was predictable as to the outcome just from one session no matter where they were in training. Now I understand what I was able to see in each individual. Thank you!
  • For those interested, a lot of the video seems to be based on two books, “Range” by David Epstein, and “Outliers” by Malcolm Gladwell. Both are good and I recommend them if you are interested in learning more about the topics in this video!