Ultralearning

“Always have a challenge,” Lewis told me as he continued with his life advice, now guiding me on a postlunch tour of central Paris: Lewis’s earlier feelings about Paris were starting to soften, and as we walked from the Notre Dame to the Louvre, his mood turned nostalgic about his days in the city. His strong opinions and passions, I would later learn, not only fueled his desire to take on ambitious challenges but could also get him into trouble. He was once detained by Brazilian federal police after an immigration officer overheard him cursing her in Portuguese to friends outside when she had denied him a visa extension. The irony was that his visa had been denied because she didn’t believe his Portuguese could be so good from such a short stay, and she suspected him of secretly trying to immigrate to Brazil outside the terms of his tourist visa.

As we continued to walk, now on the grounds in front of the Eiffel Tower, Lewis explained his approach: Start speaking the very first day. Don’t be afraid to talk to strangers. Use a phrasebook to get started; save formal study for later. Use visual mnemonics to memorize vocabulary. What struck me were not the methods but the boldness with which he applied them. While I had timidly been trying to pick up some French, worrying about saying the wrong things and being embarrassed by my insufficient vocabulary, Lewis was fearless, diving straight into conversations and setting seemingly impossible challenges for himself.

That approach had served him well. He was already fluent in Spanish, Italian, Gaelic, French, Portuguese, Esperanto, and English and had recently reached a conversational level while staying in the Czech Republic for three months. But it was his newest challenge he was planning that intrigued me the most: fluency in German after just three months.

It wasn’t, strictly speaking, Lewis’s first time with German. He had taken German classes for five years in high school and had briefly visited Germany twice before. However, like many of the students who spent time learning a language in school, he still couldn’t speak it. He admitted with embarrassment, “I couldn’t even order breakfast in German if I wanted to.” Still, the unused knowledge built up from classes taken over a decade earlier would probably make his challenge easier than starting from scratch. To compensate for the reduced difficulty, Lewis decided to raise the stakes.

Normally, he challenged himself to reach the equivalent of a B2 level in a language after three months. The B2 level—the fourth out of six levels beginning A1, A2, B1, and so on—is described by the Common European Framework of Reference for Languages (CEFR) as upper intermediate, allowing the speaker to “interact with a degree of fluency and spontaneity that makes regular interaction with native speakers quite possible without strain for either party.” However, for his German challenge, Lewis decided to go for the highest exam level offered: C2. This level represents a complete mastery of the language. To reach a C2 level, the learner must “understand with ease virtually everything heard or read” and “express themselves spontaneously, very fluently and precisely, differentiating finer shades of meaning even in the most complex situations.” The Goethe-Institut, which administers the exam, recommends at least 750 hours of instruction, not including extensive practice outside the classroom, to reach this benchmark.

A few months later, I heard back from Lewis about his project. He had missed his goal of passing the C2 exam by a hair. He had passed four of five criteria for his exam but had failed the listening comprehension section. “I spent too much time listening to the radio,” he chastised himself. “I should have done more active listening practice.” Fluency in three months of intensive practice had eluded him, although he had come surprisingly close. In the seven years after my first encounter with the Irish polyglot, he has gone on to attempt his three-month challenge in half a dozen more countries, adding to his linguistic repertoire some Arabic, Hungarian, Mandarin Chinese, Thai, American Sign Language, and even Klingon (the invented Star Trek language).

What I didn’t realize at the time but understand now was that Lewis’s accomplishments weren’t all that rare. In the space of linguistic feats alone, I have encountered hyperpolyglots who speak forty-plus languages, adventurer-anthropologists who can start speaking previously unknown languages after a few hours of exposure, and many other travelers, like Lewis, who hop from tourist visa to tourist visa, mastering new languages. I also saw that this phenomenon of aggressive self-education with incredible results wasn’t restricted to languages alone.

HOW ROGER CRAIG GAMED JEOPARDY!

“What is The Bridge on the River Kwai?” Roger Craig hastily scribbled the question on his screen. Despite first fumbling over the legibility of the film title’s final word, Craig was correct. He had won $77,000—the highest single-day winnings in Jeopardy! history at the time. Craig’s victory wasn’t a fluke. He broke records again, amassing nearly two hundred thousand dollars, the highest ever five-game winning streak. Such a feat would be remarkable on its own, but what was more incredible was how he did it. Reflecting on the moment, Craig says, “My first thought wasn’t ‘Wow, I just won seventy-seven thousand dollars.’ It was ‘Whoa, my site really worked.’”

How do you study for a test that can ask any question? That was the essential problem Craig faced as he prepared to compete. Jeopardy! is famous for stumping home audiences with trivia questions that can ask about anything from Danish kings to Damocles. Thus the great champions of Jeopardy! tend to be brainy know-it-alls who have spent a lifetime amassing the huge library of factual knowledge needed to spit out answers on any topic. Studying for Jeopardy! might feel like an impossible task, as you would need to study almost every conceivable subject. Craig’s solution, however, was to rethink the process of acquiring knowledge itself. To do that, he built a website.

“Everybody that wants to succeed at a game is going to practice the game,” Craig contends. “You can practice haphazardly, or you can practice efficiently.” To amass the wide-ranging trivia needed to break records, he decided to be ruthlessly analytical about how he acquired knowledge. A computer scientist by trade, he decided to start off by downloading the tens of thousands of questions and answers from every Jeopardy! game ever aired. He tested himself on those during his free time for months, and then, as it became clear that he was going to go on television, he switched to aggressively quizzing himself on the questions full-time. He then applied text-mining software to categorize the questions into different topics, such as art history, fashion, and science. He used data visualization to map out his strengths and weaknesses. The text-mining software separated the different topics, which he visualized as different circles. The position of any given circle on his graph showed how good he was in that topic—higher meant he knew more about that topic. The size of the circle indicated how frequent that topic was. Bigger circles were more common and thus better choices for further study. Beneath the variety and randomness in the show, he started to uncover hidden patterns. Some clues in the show are “Daily Doubles,” which allow a contestant to double his or her score, or lose it all. These extremely valuable clues may seem randomly placed, but having the entire Jeopardy! archives at his fingertips, Craig found that their position followed trends. One could hunt out the valuable Doubles by hopping between categories and focusing on high-point clues, breaking the conventional approach to the show of sticking within a single category until it was completed.

Craig also found trends within the types of questions asked. Although Jeopardy! could conceivably ask questions on any topic, the format of the game is designed to entertain a home audience, not to challenge competitors. Following this reasoning, Craig found that he could get away with studying the best-known trivia within a category, rather than digging deep into any particular direction. If a subject was specialized, he knew the answers would be geared toward the best-known examples. By analyzing his own weakness on archival questions, he could see which topics he needed to study more to be competitive. For example, he found that he was weak on fashion and focused on studying that topic more deeply.

Using analytics to figure out what to study was only the first step. From there, Craig employed spaced-repetition software to maximize his efficiency. Spaced-repetition software is an advanced flash card algorithm first developed by the Polish researcher Piotr Woźniak in the 1980s. Woźniak’s algorithm was designed to optimally time when you need to review material in order to remember it. Given a large database of facts, most people will forget what they learn first, needing to remind themselves of it again and again for it to stick. The algorithm fixes this problem by calculating the optimal time for reviewing each fact so you don’t waste energy overdrilling the same information, but also so you don’t forget what you’ve already learned. This tool allowed Craig to efficiently memorize the thousands of facts he would need for his later victory.

Although the show airs only one episode per day, Jeopardy! tapes five episodes at a time. Craig was coming back to his hotel room after winning five games straight, and he couldn’t sleep. He said, “You can simulate the game, but you can’t simulate winning two hundred thousand dollars in five hours and setting the single-day record on a game show you’ve wanted to be on since you were twelve years old.” Combining unorthodox tactics and aggressive analysis, he had gamed the game show and won.

Roger Craig wasn’t the only person I found who had seen his fortunes change as a result of aggressive self-education. I didn’t know it at the time, but in 2011, the same year my MIT Challenge would begin, Eric Barone was starting his own obsession. Unlike mine, however, his efforts would extend for nearly five years and require mastering many completely different skills.

FROM MINIMUM WAGE TO MILLIONAIRE

Eric Barone had just graduated from the University of Washington Tacoma with a degree in computer science when he thought, Now’s my chance. He had decided that he wanted to make his own video games and that now, before he got comfortable in a salaried programming job, was his opportunity to do something about it. He already had his inspiration. He wanted his game to pay homage to Harvest Moon, a charming Japanese series of games in which the player must build a successful farm: grow crops, raise animals, explore the countryside, and form relationships with other villagers. “I loved that game,” he said about his childhood experience with the title. “But it could have been so much better.” He knew that if he didn’t follow through with his own vision, that improved version would never be a reality.

Developing a commercially successful video game isn’t easy. AAA game companies budget hundreds of millions of dollars and employ thousands of people on their top titles. The talent required is similarly broad. Game development requires programming, visual art, musical composition, story writing, game design, and dozens more skills, depending on the genre and style of game developed. The breadth of skills required makes game development much harder for smaller teams than other art forms such as music, writing, or visual arts. Even highly talented independent game developers generally have to collaborate with a few people to span all the skills required. Eric Barone, however, decided to work on his game entirely alone.

Deciding to work alone came from a personal commitment to his vision and an indefatigable self-confidence that he could finish the game. “I like to have complete control over my own vision,” he explained, saying that it might have been “impossible to find people who were on the same page” regarding the design. However, that choice meant that he would need to become proficient in game programming, music composition, pixel art, sound design, and story writing. More than just a game design project, Barone’s odyssey would entail mastering each aspect of game design itself.

Pixel art was Barone’s biggest weakness. This style of art harkens back to the earlier era of video games when rendering graphics was difficult to do on slow computers. Pixel art is not done with fluid lines or photorealistic textures. Instead, a compelling image must be created by placing pixels, the colored dots that make up computer graphics, one at a time—painstaking and difficult work. A pixel artist must convey movement, emotion, and life from a grid of colored squares. Barone liked to doodle and draw, but that didn’t prepare him for the difficulty. He had to learn this skill “completely from scratch.” Getting his art skills to a commercial level wasn’t easy. “I must have done most of the artwork three to five times over,” he said. “For the character portraits, I did those at least ten times.”

Barone’s strategy was simple but effective. He practiced by working directly on the graphics he wanted to use in his game. He critiqued his own work and compared it to art he admired. “I tried to break it down scientifically,” he explained. “I would ask myself, ‘Why do I like this? Why don’t I like that?’” when looking at other artists’ work. He supplemented his own practice by reading about pixel art theory and finding tutorials that could fill gaps in his knowledge. When he encountered a difficulty in his art, he broke it down: “I asked, ‘What goal do I want to reach?’ and then ‘How might I get there?’” At some point in his work on the game, he felt his colors were too dull and boring. “I wanted the colors to pop,” he said. So he researched color theory and intensively studied other artists to see how they used colors to make things visually interesting.

Pixel art was just a single aspect Barone had to learn. He also composed all of the music for his game, redoing it from scratch more than once to make sure it met his high expectations. Whole sections of the game mechanics were developed and scrapped when they failed to meet his rigorous standards. This process of practicing directly and redoing things allowed him to get steadily better at all of the aspects of game design. Although it lengthened the time it took to complete the game, it also enabled his finished product to compete with games created by an army of specialized artists, programmers, and composers.

Throughout the five-year development process, Barone avoided seeking employment as a computer programmer. “I didn’t want to get involved in something substantial,” he said. “I wouldn’t have had the time, and I wanted to give game development my best shot.” Instead, he worked as a theater usher, earning minimum wage so that he wouldn’t get distracted. His meager earnings from his job, combined with support from his girlfriend, allowed Barone to get by as he focused on his passion.

That passion and dedication to mastery paid off. Barone released Stardew Valley in February 2016. The game quickly became a surprise hit, outselling many of the big-studio titles offered on the computer game platform Steam. Across multiple platforms, Barone estimates that within the first year of its release, Stardew Valley had sold well over 3 million copies. In months, he went from an unknown designer earning minimum wage to a millionaire named one of Forbes’ “30 Under 30” stars within game development. His dedication to mastering the skills involved played no small part in that success. Destructoid, in its review of Stardew Valley, described the artwork as “incredibly endearing and beautiful.” Barone’s commitment to his vision and aggressive self-education had paid off handsomely.

THE MIT CHALLENGE AND BEYOND

Back in my cramped apartment, I was grading my calculus exam. It was tough, but it looked as though I had passed. I was relieved, but it wasn’t a time to relax. Next Monday, I would be starting all over again, with a new course, and I still had almost a year to go.

As the calendar changed, so did my strategies. I switched from trying to do a single class in several days to spending a month doing three to four classes in parallel. I hoped that would spread the learning over a longer period of time and reduce some of the negative effects of cramming. As I made more progress, I also slowed down. My first few classes were done with aggressive haste so I could stay on schedule to meet my self-imposed deadline. After it seemed likely that I could finish, I was able to shift from studying sixty hours per week to studying thirty-five to forty. Finally, in September 2012, less than twelve months after I had begun, I finished the final class.

Completing the project was eye-opening for me. For years, I had thought the only way to learn things deeply was to push through school. Finishing this project taught me not only that this assumption was false but that this alternate path could be more fun and exciting. In university, I had often felt stifled, trying to stay awake during boring lectures, grinding through busywork assignments, forcing myself to learn things I had no interest in just to get the grade. Because this project was my own vision and design, it rarely felt painful, even if it was often challenging. The subjects felt alive and exciting, rather than stale chores to be completed. For the first time ever, I felt I could learn anything I wanted to with the right plan and effort. The possibilities were endless, and my mind was already turning toward learning something new.

Then I got a message from a friend: “You’re on the front page of Reddit, you know.” The internet had found my project, and it was generating quite a discussion. Some liked the idea but doubted its usefulness: “It’s sad that employers won’t really treat this in the same way as a degree, even if he has the same amount (or more) knowledge than a graduate does.” One user claiming to be the head of R&D for a software company disagreed: “This is the type of person I want. I really do not care if you have a degree or not.” The debate raged. Had I actually done it or not? Would I be able to get a job as a programmer after this? Why try to do this in a year? Was I crazy?

The initial surge of attention led to other requests. An employee at Microsoft wanted to set me up for a job interview. A new startup asked me to join its team. A publishing house in China offered me a book deal to share some studying tips with beleaguered Chinese students. However, those weren’t the reasons I had done the project. I was already happy working as a writer online, which had supported me financially throughout my project and would continue to do so afterward. My goal for the project wasn’t to get a job but to see what was possible. After just a few months of finishing my first big project, ideas for new ones were already bubbling up inside my head.

I thought of Benny Lewis, my first example in this strange world of intense self-education. Following his advice, I had eventually reached an intermediate level of French. It had been hard work, and I was proud that I had been able to push against my initial difficulty of being surrounded by a bubble of English speakers to learn enough French to get by. However, after finishing my MIT experience, I was injected with a new confidence I hadn’t had in France. What if I didn’t make the mistake I made last time? What if, instead of forming a group of English-speaking friends and struggling to pop out of that bubble once my French was good enough, I emulated Benny Lewis and dived straight into immersion from the very first day? How much better could I be, if as in my MIT Challenge, I held nothing back and optimized everything around learning a new language as intensely and effectively as possible?

As luck would have it, around that time my roommate was planning on going back to grad school and wanted some time off to travel first. We’d both been saving, and if we pooled our resources and were frugal in how we planned our trip, we figured we might be able to do something exciting. I told him about my experiences in France, both of learning French and of secretly believing that much more was possible. I told him about the social bubble that had formed when I had arrived without speaking the language and how difficult it had been to break out of it later. What if, instead of just hoping you’d practice enough, you don’t give yourself an escape route? What if you commit to speaking only the language you’re trying to learn from the first moment you step off the plane? My friend was skeptical. He had seen me study MIT classes for a year from across our apartment. My sanity was still an open question, but he wasn’t as confident in his own ability. He wasn’t sure he could do it, although he was willing to give it a shot, as long as I didn’t have any expectations of him to succeed.

That project, which my friend and I titled “The Year Without English,” was simple. We’d go to four countries, three months each. The plan in each country was straightforward: no speaking English, either with each other or with anyone we’d meet, from the first day. From there we’d see how much we could learn before our tourist visas ran out and we were pushed to a new destination.

Our first stop was Valencia, Spain. We had just landed in the airport when we encountered our first obstacle. Two attractive British girls came up to us, asking for directions. We looked at each other and awkwardly sputtered out the little Spanish we knew, pretending we didn’t speak any English. They didn’t understand us and asked us again, now in an exasperated tone. We stumbled over some more Spanish and, believing we couldn’t speak English, they walked away in frustration. Already, it seemed, not speaking English was having unintended consequences. Despite that inauspicious beginning, our Spanish ability grew even faster than I had anticipated. After two months in Spain, we were interacting in Spanish beyond what I had achieved in an entire year of partial immersion in France. We would go to our tutor in the morning, study a little at home, and spend the rest of the day hanging out with friends, chatting at restaurants, and soaking up the Spanish sun. My friend, despite his earlier doubts, was also a convert to this new approach to learning things. Although he didn’t care to study grammar and vocabulary as aggressively as I did, by the end of our stay, he too was integrating seamlessly into life in Spain. The method worked far better than we had hoped, and we were now believers.

We continued the trip, going to Brazil to learn Portuguese, China to learn Mandarin, and South Korea to learn Korean. Asia proved a far harder task than Spain or Brazil. In our preparation, we had assumed those languages would be only a little more difficult than the European ones, although it turned out that they were much harder. As a result, our no-English rule was starting to crack, although we still applied it as much as we could. Even if our Mandarin and Korean didn’t reach the same level of ability after a short stay, it was still enough to make friends, travel, and converse with people on a variety of topics. At the end of our year, we could confidently say we spoke four new languages.

Having seen the same approach work for academic computer science and language-learning adventures, I was slowly becoming convinced that it could be applied to much more. I had enjoyed drawing as a kid, but like most people’s attempts, any faces I drew looked awkward and artificial. I had always admired people who could quickly sketch a likeness, whether it be street-side caricaturists to professional portrait painters. I wondered if the same approach to learning MIT classes and languages could also apply to art.

I decided to spend a month improving my ability to draw faces. My main difficulty, I realized, was in placing the facial features properly. A common mistake when drawing faces, for instance, is putting the eyes too far up the head. Most people think they sit in the top two-thirds of the head. In truth, they’re more typically halfway between the top of the head and the chin. To overcome these and other biases, I did sketches based on pictures. Then I would take a photo of the sketch with my phone and overlay the original image on top of my drawing. Making the photo semitransparent allowed me to see immediately whether the head was too narrow or wide, the lips too low or too high or whether I had put the eyes in the right spot. I did this hundreds of times, employing the same rapid feedback strategies that had served me well with MIT classes. Applying this and other strategies, I was able to get a lot better at drawing portraits in a short period of time (see below).