Ultralearning

by Scott Young

  In the face of this, one must simultaneously have confidence and deep humility. Without the belief that progress in one’s own knowledge and skill is possible, one cannot undertake the project required to generate it. This kind of confidence may be mistaken for arrogance by outsiders, as it can seem that an effort to learn something quickly and intensely is somehow an assertion that the subject is trivial or that, having learned something, one has learned everything. Instead, this confidence must be paired with deep humility. In every project I’ve undertaken, including this one, my thoughts upon concluding it were not to think I had finished but to suddenly become aware of how much further I could have gone. Before I started my MIT Challenge, I imagined that covering an undergraduate degree’s worth of computer science concepts would be plenty. After I had finished, I could see how each topic I had learned could be multiplied into a doctorate’s worth of research or a lifetime spent coding to fully understand it. My experience in learning languages to a level where I could hold conversations made me realize how many more words, expressions, nuances of culture, and difficult communication situations were left to explore. Finishing a project, therefore, isn’t usually accompanied by a sense of finishing learning but by the creation of a feeling of possibility as one’s eyes are opened to all the things left to learn.

  It’s this aspect of learning that I find most interesting. Many pursuits in life have a kind of saturation point, after which the longing for more of a thing eventually diminishes as you get more of it. A hungry person can eat only so much food. A lonely person can have only so much companionship. Curiosity doesn’t work this way. The more one learns, the greater the craving to learn more. The better one gets, the more one recognizes how much better one could become. If you finish reading this book and have been encouraged to try your own project, this would be my greatest hope—not that you’d be successful at your project but that your ending would be a beginning. That by opening a small crack in all the possibly knowable things there are in the world, you might peer through and find there is far, far more than you had ever imagined.

  Acknowledgments

  This book could not have happened without the help, advice, and work provided by many different people. First, I would like to thank Calvin Newport. Had it not been for his early encouragement, I might never have pursued writing a book about this topic. I would also like to thank Benny Lewis, whose early inspiration and endless advice over the years has had such a strong influence on my thoughts on learning and writing. Laurie Abkemeier, my agent, was instrumental in taking my rough ideas in a proposal and pushing me to develop something worthy of print. I thank Stephanie Hitchcock for editing the book and providing me with excellent feedback and suggestions. I’m also thankful to my friends and family who read over early drafts of the proposal and manuscript, helping the idea take shape. In particular, I would like to thank Zorica Tomovska, Vatsal Jaiswal, Tristan de Montebello, James Clear, Josh Kaufmann, Kalid Azad, and Barbara Oakley for their early feedback.

  I would like to thank the wonderful people I met and interviewed while preparing for the book. I am grateful to Roger Craig, Eric Barone, Vishal Maini, Diana Jaunzeikare, Colby Durant, and Vatsal Jaiswal, who were kind enough to take time to help me fill in the details of their incredible stories. I want to thank many of the researchers who walked me through their findings and helped me understand the science of learning better. In particular, I want to thank K. Anders Ericsson for his patience as he helped me clarify many important points. In addition, I thank Robert Pool, Jeffrey Karpicke, Angelo DeNisi, Avraham Kluger, Jacqueline Thomas, and Michael Herzog for helping me understand the nuances of the science discussed in this book. I want to thank all the people who participated in my experiments with coaching ultralearning: Tristan de Montebello, Jeff Russell, Diana Fehsenfeld, Kate Schutt, Lissa Sherron, Joshua Sandeman, Keerthi Vemulapalli, Brittany Hsu, Shankar Satish, Ashima Panjwani, Ashfaq Alsam, Deepti Kannapan, and Ankita J.

  Finally, I want to thank my parents, Douglas and Marian Young, both teachers, who taught me that learning is its own reward.

  Appendix

  Further Notes on My Ultralearning Projects

  The MIT Challenge

  Goal: Learn the material taught in MIT’s undergraduate curriculum for computer science, using their freely provided materials and used textbooks

  Method: Aim to pass all of the final exams (score over 50 percent, unless other information was provided) and complete the programming projects

  Timeframe: October 2011 to September 2012

  Notes and Discussion

  It’s important to note that what I ended up completing was not a facsimile of an MIT degree. Although I strove, whenever possible, to benchmark the overall curriculum covered and the intensity of evaluation, there were necessary departures from how an actual MIT student would have progressed through the same material.

  At the level of the entire curriculum there were changes. MIT’s OpenCourseWare didn’t offer options for humanities that I could grade myself at the time, so I swapped those for classes in economics. Lab-heavy classes for which I didn’t have access to the equipment were substituted for pencil-and-paper theory classes I could do. MIT students were expected to carry out a thesis project. I didn’t do that during my twelve-month study period, but for fun I did create a computer program that would allow someone to play Scrabble against a computer opponent shortly after my project had officially concluded. In evaluating the programming projects, I simply counted them as a success if they worked and performed the desired functions or were able to complete the accompanied testing suites.

  For final exams, my default benchmark was to achieve at least 50 percent. I stuck to the official grading rubric whenever possible. When there were gaps (such as how to deduct points for arithmetic or algebraic errors on multistep problems), I used my judgment. The latter stage introduced some potential bias, so I decided to go back several years after completing the challenge and reevaluate all my exams using the strictest possible grading scheme (any mistake on a multipart question would make the entire question worth zero points; any incorrect result applied to further questions would make those questions also worth zero points). The outcome was that six of the thirty-three classes I had recorded as a “pass” would have been counted as a “fail” under this stricter schema. I don’t believe this evaluation is the correct one, so I stand by my original evaluation of having passed those exams, but it is worth pointing out to show how much impact my subjective decisions had. A few classes had no final exams, so in those cases evaluation defaulted to assignments or midterm exams. Completing assignments was not a requirement to complete a class, however I did end up doing many of them as part of the learning process.

  For more information on the challenge, such as course lists, materials used and scans of my exams, you can visit the challenge homepage: www.scotthyoung.com/blog/mit-challenge/.

  The Year Without English

  Goal: Learn Spanish, Portuguese, Mandarin Chinese, and Korean

  Method: Avoid speaking English for the entire year, while traveling to Spain, Brazil, China, and South Korea (roughly three months each). I did this project along with Vatsal Jaiswal (who is also mentioned in chapter six).

  Timeframe: September 2013 to August 2014

  Notes and Discussion

  Quantifying the level of proficiency reached in each language is a tricky task. There’s a dual risk, both of exaggeration—implying a perfect level of fluency when that is probably a process requiring decades of immersion—and of downplaying. One person I talked to after the trip asked if I could “give a taxi driver directions” even though this task only requires a few hours of practice, not months. So with those difficulties in mind, I’ll try to estimate the level that we reached:

  Spanish: Here I believe both my friend and I reached a roughly B2 level after three months, meeting Benny Lewis’s standard for fluency (although certainly not everyone’s standard). At that level, we had little difficulty socializing fo
r hours on any topic in Spanish, although certainly our accent, grammar and more formal speaking abilities were not at the level of a native speaker.

  Portuguese: We were weaker in Portuguese than in Spanish, although not substantially so. The two languages share a common base, so there was much less to learn than there had been with Spanish. We could make friends and socializing, but not quite as effortlessly.

  Mandarin Chinese: This marked the first big divergence in our abilities. I had really wanted to learn Chinese and had spent some time on flash cards prior to our trip to familiarize myself. My friend was less interested and struggled more. In the end, I wrote and passed the HSK 4 exam (the fourth in a six-level series of exams measuring Chinese proficiency) and I would say my Mandarin was decent, although more limited on advanced topics, where the vocabulary is completely different from English. My friend reached a lower-intermediate level, being able to speak comfortably and use tones but with less vocabulary.

  Korean: In this language we both reached a lower-intermediate level, able to have conversations and get by in daily life, but on a more restricted range of topics. Part of this was the difficulty of the Korean language, but a bigger part was simply that as it was the fourth new language in a row, we were getting burned out.

  Although we aimed to do most of our learning after arriving in each country, we did do some prior preparation for each. This was mostly listening to Pimsleur audio tapes and doing some flash cards. In generally we spent around twenty-five to fifty hours per language, although I spent more on Chinese (approximately one hundred hours) prior to arriving.

  Those interested can see more about our project (including videos we put together showing our progress in each country), what we used to learn as well as unscripted interview to show roughly the level we reached in each language on the project homepage: www.scotthyoung.com/blog/the-year-without-english/.

  Portrait Drawing Challenge

  Goal: To improve my ability to draw faces realistically

  Method: Rapid feedback, techniques from various books and courses

  Timeframe: July 2016

  Notes and Discussion

  This was a shorter project, taking one month and totaling one hundred hours of practice. In addition to the strategy of drawing quick sketches and comparing them by overlaying them on semi-transparent reference photos, I also greatly benefitted from the book Drawing on the Right Side of the Brain, and from Vitruvian Studio’s Portrait Drawing class.

  I’ve uploaded every drawing, sketch, and self-portrait I did, along with a more detailed discussion of what I used to learn on the project homepage: www.scotthyoung.com/blog/myprojects/portrait-challenge/.

  Further Challenges

  At the time of writing this book, the above three challenges are my main public ultralearning projects. However, I’m always learning new things, so as I do more public challenges, I’ll post them here: www.scotthyoung.com/blog/my-projects/.

  Notes

  Chapter I: Can You Get an MIT Education Without Going to MIT?

  1. The Goethe-Institut, which administers: “Further Information,” Goethe-Institut, https://www.goethe.de/en/spr/kup/prf/prf/gc2/inf.html.

  2. “My first thought wasn’t ‘Wow’”: Thanh Huynh, Roger Craig—Knowledge Tracking, filmed August 2011, YouTube video, 14:20, posted November 2011, https://www.youtube.com/watch?v=jmld3pcKYYA&t=1s.

  3. “Everybody that wants to succeed at a game”: “How One Man Played ‘Moneyball’ with ‘Jeopardy!,’” National Public Radio, https://www.npr.org/2011/11/20/142569472/how-one-man-played-moneyball-with-jeopardy.

  4. Spaced-repetition software is: Gary Wolf, “Want to Remember Everything You’ll Ever Learn? Surrender to This Algorithm,” Wired, April 20, 2008, https://www.wired.com/2008/04/ff-wozniak/?currentPage=all.

  5. “You can simulate the game”: Huynh, Roger Craig—Knowledge Tracking.

  6. “incredibly endearing and beautiful”: Patrick Hancock, “Review: Stardew Valley,” Destructoid, March 7, 2016, https://www.destructoid.com/review-stardew-valley-345495.phtml.

  7. “This is the type of person”: “College Too Expensive? This Guy Just Finished a Four Year Computer Science Program in ONE Year Using Free MIT Material” (video), Reddit, https://www.reddit.com/r/videos/comments/10tk9j/college_too_expensive_this_guy_just_finished_a/.

  8. Done without the benefit: Steve Pavlina, “Graduating College in 3 Semesters,” December 4, 2005, https://www.stevepavlina.com/blog/2005/12/graduating-college-in-3-semesters/.

  9. Diana Jaunzeikare embarked on: Diana Jaunzeikare, “Personal PhD.” https://diana.is/personal-phd.

  10. “70–80+ hours each week”: Tamu, “Independent Chinese Study: Review,” Chinese-forums.com, https://www.chinese-forums.com/forums/topic/43939-independent-chinese-study-review/.

  11. Trent Fowler, starting in early 2016: Trent Fowler, The STEMpunk Project (Self-published, 2017).

  Chapter II: Why Ultralearning Matters

  1. “Average is over”: Tyler Cowen, Average Is Over: Powering America Beyond the Age of the Great Stagnation (New York: Penguin, 2013).

  2. The MIT economist David Autor: David H. Autor, Lawrence F. Katz, and Melissa S. Kearney, “The Polarization of the U.S. Labor Market,” American Economic Review 96, no. 2 (May 2006): 189–94.

  3. Tuition has increased far faster: Danielle Douglas-Gabriel, “College Costs Rising Faster than Financial Aid, Report Says,” Washington Post, October 26, 2016,https://www.washingtonpost.com/news/grade-point/wp/2016/10/26/college-costs-rising-faster-than-financial-aid-report-says/?utm_term=.72c95b4c86cb.

  4. “a leading English-language novelist”: Gareth Cook, “The Singular Mind of Terry Tao,” New York Times, July 24, 2015, https://www.nytimes.com/2015/07/26/magazine/the-singular-mind-of-terry-tao.html.

  Chapter IV: Principle 1—Metalearning: First Draw a Map

  1. “Kuti paoka djalou”: Linguistic Society of America, “‘Monolingual Fieldwork Demonstration’—Daniel Everett,” filmed July 2013, YouTube video, 1:16:27, posted September 2013, https://www.youtube.com/watch?v=sYpWp7g7XWU.

  2. What makes this feat particularly impressive: To avoid ruining the demonstration by using English, a language the other speaker might have been familiar with, Everett chose to phrase all his initial queries in the Pirahã language, spoken only by a remote people in the Amazon jungle of Brazil.

  3. Over the last thirty years: The unusualness of this language has led to somewhat of a controversy in linguistics, with Dan Everett’s claims about Pirahã’s grammar at center stage in an attack on linguistic orthodoxy.

  4. To see why metalearning is so: Jacqueline Thomas, “The Role Played by Metalinguistic Awareness in Second and Third Language Learning,” Journal of Multilingual and Multicultural Development 9, no. 3 (1988): 235–46, https://www.tandfonline.com/doi/abs/10.1080/01434632.1988.9994334.

  5. Determine if learning: Don’t take this to mean that I think grad school is useless. The important thing to decide is whether it will really matter to you, depending on the job you want, the subject of your study, and the institution. My point isn’t that grad school is a waste of time but rather that when making a decision involving so much time and cost, you’d better do the research first!

  6. For example, one common recommendation: Victor Mair, “How to Learn Chinese and Japanese,” Language Log, February 17, 2014, http://languagelog.ldc.upenn.edu/nll/?p=10554.

  7. The literature on self-directed learning: George E. Spear and Donald W. Mocker, “The Organizing Circumstance: Environmental Determinants in Self-Directed Learning,” Adult Education Quarterly 35, no. 1 (March 1, 1984): 1–10, https://journals.sagepub.com/doi/abs/10.1177/0001848184035001001?journalCode=aeqb.

  8. That led me to do: “Portrait Drawing—The Complete Online Course,” Vitruvian Studio, https://vitruvianstudio.com/course/portrait-drawing/.

  Chapter V: Principle 2—Focus: Sharpen Your Knife

  1. Somerville explained, “she would have been contented”: Mary Somerville, Personal Recolle
ctions, from Early Life to Old Age, of Mary Somerville: With Selections from Her Correspondence (London: Roberts Brothers, 1874), 23.

  2. Hence, skilled performers: K. Anders Ericsson, The Road to Excellence: The Acquisition of Expert Performance in the Arts and Sciences, Sports, and Games (New York: Psychology Press, 2014), 25.

  3. Similarly, the phenomenon: John Dunlosky, Katherine A. Rawson, Elizabeth J. Marsh, et al., “Improving Students’ Learning with Effective Learning Techniques,” Psychological Science in the Public Interest 14, no. 1 (January 8, 2013): 4–58, https://journals.sagepub.com/doi/abs/10.11771529100612453266.