It’s the end of the year, let me talk about myself.

The title says it all.

No, seriously, this post is all about me. So much for not talking about me…

This may be cliché but, on this very last day of 2018, I want to end the year with a reflection of things that have greatly influenced me this year. These “things” come in various shapes and forms: persons, books, podcasts, courses, events, and YouTube series. Whatever they are, as long as they left a big impression on me and helped me grow, be it personally or professionally, I’m going to list them here. I decided to do this because I realized that I need more documentations in my life to be more conscious of what I have done, what I’m still missing, and where I am headed. In other words, I need more introspections and writing just always makes things clearer for me. Hence, this is really an exercise for myself. That said, if you happen to find any of the things I listed below helpful, that would be my great pleasure too.

I’m going to organize this list of things by their influence on my career, my personal growth, and my views on the society. As the section headers suggest, this is really a hodgepot of topics ranging from technical development to my own personal life and viewpoints. I have decided to include all of them here in a supposedly technical blog because my work is only a small part of me and is an even smaller part that I had identified needed working on this year. To me, 2018 has been a year of changes and I want to honestly present here all that I have learned through them.

Career

Late last year, being ambitious and overly confident, I switched to a team that focuses on NLP and deep learning, only to find that I was ill-prepared. Surrounded by literally field experts, I had never felt more like an impostor. I remember even feeling intimidated when interviewing candidates, fearing I would say something stupid that would tarnish their opinions on the team. Fortunately, by this time, I was so familiar and comfortable with the impostor syndrome that none of this felt new or shocking to me.¹ To quickly catch up in this field, I consulted many resources and the following few worked particularly well for me.

Andrew Ng’s deep learning course on Coursera

Like many other beneficiaries, I particularly appreciate Ng’s lucid and vivid explanations to complex subjects that made them no longer inaccessible. For example, I still remember the “eureka” moment I had after watching his video on GRU and LSTM, a concept that I had at best only roughly grasped before. Another thing that I appreciate a lot is his programming assignments, which are all written in Python and Keras and hosted in Jupyter Notebooks. In particular, I find great reuse values in these notebooks and find myself using them as base frameworks to develop my solutions on top of over and over again.² For example, the exercise on seq2seq model with attention is particularly helpful to me because it’s the most readable Keras implementation that I have found.³ To summarize, I find this course excellent in developing the right intuitions for the common deep learning algorithms, which makes it easier to learn more advanced concepts later on.

fast.ai

fast.ai is an organization founded by Jeremy Howard and Rachel Thomas, which, among other things, offers a completely free MOOC on deep learning. Many students appreciate the top-down approach which teaches them to build a model first before gradually explaining all the theories behind. Personally, I’m not a big fan of this approach because I find it frustrating not knowing exactly what I am doing. However, I am a big fan of their philosophy that there is absolutely nothing complex about deep learning or AI, and anyone can understand it if being explained to in a clear and intuitive manner. As a self-taught machine learning practitioner, this resonates extremely well with me.⁴ This philosophy is evident throughout their teaching. A particularly memorable lesson to me is when Jeremy took a paper and showed us step by step how he would go on to read it, especially how to translate a seemingly foreign concept into something that we have in fact learned in a different form before. Truthfully, coming outside academia, reading papers had always appeared to be an intimidating task for me. Seeing how Jeremy approached and dissected it really opened my eyes and provided me with enough confidence to perform my future research independently.

EMNLP

Held in Brussels, Belgium, the 2018 EMNLP is the first NLP conference that I have ever attended.⁵ Needless to say, my impostor syndrome was at its peak again. Also since I had no paper submitted, my goal in the beginning was just to attend talks and try to learn as much as possible. To be honest, this is not a very bad strategy as many talks did introduce me to new methods and directions and I came back with many fresh ideas to try for my own projects. However, I felt like a dutiful student, rather disconnected from the speakers and other attendees. Fortunately, this changed during the social event where, with free food and alcohol, I got to mingle with many of them, learned about their research interests and processes, and was genuinely surprised to find that they were interested in hearing what I do with NLP at work too.⁶ To me, this just goes on to show the importance of staying curious and open-minded, and the stupidity of fearing to look stupid.⁷ In the end, I walked away with a rather fulfilling experience and ready for more conferences in the coming year.

DataKind

In the second half of this year, I had a mini existential crisis where I can no longer identify the societal impact of my work. This is also the year where the general public, including me, lost trust in the tech industry and questioned the real values that all the silicon valley companies bring to the society. Am I really helping the world and its people or am I just helping the rich get richer while providing marginally incremental values to some middle-class users of our app? Surely, the work is interesting and I feel intellectually challenged and fulfilled, but is this really all I need to get myself out of bed every morning? I determined that it is not, and proceeded to research on the various “data science for good” opportunities. In the end, I reconnected with DataKind, a nonprofit organization that connects volunteering data scientists with nonprofits and small for-good companies who cannot afford data scientists themselves. I first discovered DataKind more than three years ago when I participated in a weekend-long DataDive project for a local nonprofit with like-minded fellow data scientists. Although the experience left a big impression on me, I had not since followed up with more volunteering. Three years later, seeing DataKind still at the forefront of this laudable effort, it only seemed natural for me to reconnect with the organization and assume a bigger role this time. Hence, a couple months ago, I joined the SF chapter as a data embassador and have been since working with a new nonprofit called Empower Work that provides free SMS-based counseling to people going through a difficult situation at work (e.g., stress, bias, harassment). Although at this early stage, my involvement is still limited, I have been nevertheless savoring the small joy of meeting and talking with the people from the organization, trying to understand their data problems, and working with other volunteers to offer our advice. An important lesson that I have learned through the experience is to never doubt the amount of contributions you can bring to the table as long as you are willing to contribute. When the SF chapter lead asked me to spearhead this project given my NLP background, I felt unease at first and was unsure of my capabilities and the outcome. Now, after having successfully hosted a project accelerator event, crystallized the next steps, and received the organization’s thank-you note, I’m more than glad that I did it. Finally, to quote Paul Duan, the founder of another data science nonprofit called Bayes Impact, data science is after all my strongest lever to do some good to the society, and it would be incredibly wasteful to only use it to increase the click-through rate or some kind of user engagement metrics.

Personal growth

Now we are getting personal. If you read my post from the beginning, you would recognize the impostor syndrome as a repeating theme so far. For the past few years, feeling perpetually left behind, my new year’s resolution had always been to challenge myself. Starting this year, I have changed it to being comfortable with being challenged, and it turned out to be harder said than done. Without having properly managed the stress and self-doubt that come with each challenge, I ended up feeling overwhelmed and burned out. At the same time, I had some other issues that I had never gotten to think deeply about, which in turn made this inferiority complex worse. Lessons learned: don’t ignore your feelings. In the first half of the year, I tried various ways of dealing with these problems, including meditation, running, and early morning routines, only to realize that by doing so, I was not, in fact, dealing with anything because, after all, the goal of these strategies is to be at ease with one’s feelings. For me, however, I can only move on from a problem if I actually solve it. With that, the following are a couple resources that have helped me find my solutions.

The School of Life

The School of Life is a company that aims to provide people with resources on emotional intelligence. One of such resources is a rich YouTube channel filled with brightly animated self-help videos on issues ranging from depression and anxiety to love and relationships, all narrated with a perfect British accent.⁸ My favorite video is this one called the importance of a nervous breakdown, which, as the name suggests, normalizes the experience of having a breakdown and depression, and goes on to argue that it is in fact a beneficial experience as it forces us to face and sift through the multitude of problems that we had, over the years, hastily swept under the carpet because we were convinced that we had no time to deal with them and that normal and successful people don’t have them. This latter notion, that everyone around me is better put together and that I should thereby not alarm them with my faults and oddities, hits particularly home for me. I especially liked this parting message from the video: “our crisis, if we can get through it, is an attempt to dislodge us from a toxic status quo and it represents an insistent call to rebuild our lives on a more authentic and sincere basis.” In other words, we need to be honest with ourselves. Outside the realm of personal growth, this discussion reminds me of a similar debate that people in the tech industry are having about its culture. In an industry where everyone is supposedly “winners” and overachievers, admitting failures, or even just mistakes, is often advised against or at least needs to be sugar-coated. In my opinion, this is incredibly stupid and toxic, and sets a company to its own nervous breakdown. After seeing this video and having these reflections, one thing that I started to do at work is to be forthcoming when reporting and sharing my own failures (because what I do at work and in life need to be in sync). No sugarcoating, no blaming others. Simply, I fucked up. Here is how I did it wrong, and here is how I’ll do it better. It’s liberating.

Nouvelle École

This is a podcast in French that I discovered while learning the language.⁹ It is hosted by a 25-year old guy called Antonin Archer who interviews all the people that he personally admires. These people include authors, chefs, artists, singers, wrestlers, and pro-gamers. What initially drew me to this podcast is Antonin’s incredible curiosity and open mindedness, evidenced in the diverse group of people that he invites to the program. Even though all these people are successful in their own way, many of them lead a very different life from his. Despite that, he never judges, never assumes, but always listens, and always remains curious. To me, that is one of the highest virtus and is the key to mutual understandings. Often in life, we are content with socializing with people from the same industry or the same social class and rarely venture outside this immediate circle, which I have come to realize is a big mistake. I used to blame this on the city of San Francisco that is highly concentrated with tech workers like me. Surely, if I move to a diverse city like Paris, I will be more open-minded and befriend a diverse bunch of people like Antonin, won’t I? Truthfully, I doubt it since, psychologically, this is still something that I need to work on. I once read a person’s goal being talking to at least X new people in every party he/she attends. To this, despite being already a challenge to me, I’ll go on to add the constraint of talking to people coming from X different backgrounds instead.

Aside from the diversity of the guests, despite the free style of the interview, Antonin is also good at extracting the key, savory messages from the free-flowing conversations. Some of my favorite episodes include this star dancer from the Paris Opera Ballet who, since the age of 8, determined to be a ballet dancer and finally obtained the nomination of the “star” title once she was no longer purposefully trying and letting herself simply enjoy the performance and the art, this comedian / author who talks freely about her fear, her doubts, her solitudes, and the importance of being spontaneous, and this journalist who dedicates her career to listening to and understanding people and helping them share their stories with the world while helping the world hearing each other. These stories made me realize that there are many different ways of leading a fulfilling life and there is really no reason to ever feel stuck. If you do, don’t sit around waiting for other people or events to unblock you. Unblock yourself.

My views on society

Alright, enough talking about me. Driven by the worldwide political and social events, I have, for the first time, taken a strong interest in all the societal issues in 2018. The question that I set off to answer is 1) what constitutes a good society and 2) how we can make one. Needless to say, I ended up with more questions than answers.

21 Lessons for the 21st Century

This is the last book that I read in the year and, in my opinion, summarizes and supplements the other books that I have read in the same vein very well. Before talking about this book, I’d like to describe another one that I had read earlier and enjoyed a lot. It’s Utopia for Realists: How We Can Build the Ideal World by Rutger Bregman. The book has three main ideas: universal basic income, 15-hour work week, and open borders, among which, the first and the last resonated me the most. Many would agree that a good society is an equal society - just look at the Scandinavian countries. But these countries either has a big government with socialist values that are at odds with the capitalism that most western countries adopt, or it sits on top of some incredibly rich natural resources that other countries can only dream for. So how do we reduce existing inequality among countries that, to favor economic growth, widened the wealth gap to incentivize some people to get rich first? According to Bregman, the answer is to give everyone free money, hence the so-called universal basic income. I particularly liked the experiment that he talked about in the beginning of the book where, in 2009, the British government gave £3,000 to each of the 13 homeless people living on the street of London, who, in total, had cost the government £400,000 a year in the name of homeless management, so £3,000 for 13 persons by itself translate into a saving of more than 10 folds. The money was given with no strings attached, and at the end of the year, the government found that, not only did the money not yet entirely spent, 11 of the 13 people have successfully left the street and moved to the apartments. If the experiment sounds like an outlier, Bregman also presented similar experiments conducted in Kenya, in Canada, and even in the US, all with the same conclusion, that is, giving people free money is the most effective way to reduce poverty because, despite popular bias against poor people, they themselves know best how to spend them in order to escape the poverty trap.

Harari, on the other hand, is more cool-headed and naturally sceptical about everything. One of his key messages in 21 Lessons for the 21st Century is that massive advances in AI and biotechnologies are coming to us in strides and our society is very ill-prepared. Above all, what should we do with the newly created “useless class” who, despite being able-bodied, find themselves unemployable because of machines and algorithms? Every government knows a high unemployment rate links inevitably to a high level of social unrest, so what can one do to prevent it? Following Bregman’s thoughts, the logical next step should be to implement universal basic income to make sure everyone has a safety net to fall back on, but Harari argues that it is far easier said than done. For example, he challenges the notion of “universal” and points out that, realistically, such a system can only be implemented on a national basis and cannot be truly universal. Okay, but what are the nations that can reasonably afford it in such a big scale? Harari soberly warns that those that can are those that don’t need them in the first place. So what about those that actually needs the free money for their citizens? No idea. For example, if the first wave of automation designed by the tech giants in the US renders a whole class of works in the third-world countries useless, does US have the obligations to offer any basic income to them? Morally speaking, yes, but would any government actually be willing to do that? Very doubtful. Harari also challenges the notion of “basic” and predicts that there would be no agreed-upon definition as to how much is basic enough. In other words, trying to implement such a system, however sound it is in theory, would be nothing but an execution nightmare and it is very likely that it won’t be ready by the time the robots are.

It should be noted that this book is not a collection of solutions but that of questions, and it seems to me that the goal of Harari is simply to get more people worried so he can take a breath. Well played.

Seriously, I do believe that, in the tech industry, the very place where all these advances and challenges are being produced, there are dangerously too few people who think about the impact of what they are doing. If you think you are simply writing code, think deeper. There is no such thing as “simply writing code” as every line of code written completes something bigger, which in turn snowballs into an even bigger, and sometimes longlasting, impact on the society. Once again, the question is how we limit such an impact. To that, I have no idea, but I agree with Harari that once enough smart people, many of who serve as the brain to these tech giants, start to think about it, we may just well have our solution.¹⁰

Immigration

This year, driven by many news stories, I have spent a lot of time thinking about immigration, an issue that is particularly dear to me being an immigrant myself. A particular topic that’s especially fascinating to me is how the US, a country that was built by immigrants and publicly supports multiculturalism but paradoxically has a long and violent history on racism, treats and thinks about its immigrants today, and how that’s different from a country like France, who swears by law that everyone is equal and that even the very notion of race does not exist, but goes on to consolidate this belief by demanding all minorities to behave just like the French. This summer, after the French team won the world cup, a seemingly harmless joke triggered a heated international debate on this very topic. Specifically, drawing upon the origins of the French team members, American comedian Trevor Noah made a joke saying that it is in fact Africa that won the world cup. This rather light-hearted comment sparked an uproar across the ocean in France with people of all classes taking offense and calling Noah out for being racist. What is going on?

As many international conflicts, the problem comes from cultural misunderstandings. In the US, it is a norm to identify people with their origins and calling each other African American, Asian American, and so on. In fact, Noah, who is of African origin himself, believed that, by properly attributing these players to their origin, he is collecting with them and showing his pride. However, in France, calling people African or Asian is equivalent to denying them of their “Frenchness” and alluding them being inferior to the other citizens, which is essentially being racist. If this sounds strange, you need to remember that the two countries adopt two fundamentally different immigration models. In the US, at least on paper, we welcome and embrace diversity and tolerate communitarianism. In France, however, people believe in integration and assimilation, which is to say they expect immigrants, once having landed in France, to quickly learn the French language and discretely integrate into the French society and assume the French identity. In other words, whoever you were before no longer matters as you are now simply French. This is why by calling them African, Noah inadvertently crossed the line and unknowingly insulted them.

When learning about these differences, as an immigrant myself, my first impression is that the French model is extremely unfair to the immigrants because, by forcing the minority to lose their identity and quickly adopt a brand new culture, it is on the other hand making it extremely easy for the majority to cohabit with them, all while assuming a moral high ground of allowing everyone to be equal. The American model, on the other hand, appears to be so liberal and ideal. Essentially it’s like saying: “Welcome to America. You don’t need to change a thing.” Furthermore, I believe a society that embraces diversity, once passing the initial unease, is simply stronger in the long run. Hence, I sided naturally with the Americans.

But does it actually work in reality? Or if it does, how does one explain the persistent police brutality, the white supremacy, and the scapegoating of immigrants for the unemployment? If you told us to be proud of our differences, why do you turn around and use them against us? All these inconsistencies made me question the realism of the American ideology and postulate that, if you really want people to see each other equally, maybe you shouldn’t emphasize so much on our differences. Wait, isn’t that what the French immigration system based on?

Okay, switching sides. Let’s see if the French way works. On the surface, it does appear so. When Noah made that now infamous joke, it seems that everyone got offended, including the players themselves, because deep down, they do believe they are all equal and they are all French. By forcing everyone to be the same and training the citizens to be colorblind do seem to be an effective strategy. But are people really equal? It turns out, once again, not really. Because the French law prohibits the collection of any data based on race (as doing so would be racist), my understanding of this problem comes mainly from this podcast called Kiffe ta race where people of racial minorities come to share their experiences of living in France today. According to them, the discrimination still exists and the dogmatic emphasis on universality makes it extremely hard to talk about it, let alone having any data to support it. Given that, maybe the American model isn’t so bad after all?

Truthfully, to take a step back, the two countries are not that comparable to start with. Among other things, the US had its long, ugly history of slavery right at its doorstep while France had its colonization overseas, which makes it easier for them to turn the page. That said, all things considered, I still believe the American ideology in multiculturalism is the long-term solution here but to realize it, our work is far from being done. As summarized beautifully by this series of tweets written by Noah Smith, “a country with institutions strong enough not to have to rely on homogeneity will be the strongest country imaginable.” To me, these moonshot goals are what makes America still great, in which people strive for the best and don’t get caught up by falling short. To me, these are the real American dreams.

Conclusions and what’s next

Okay, before getting too carried away on New Year’s Eve, let me end this post by saying that, as much as I have learned in the past year, there are still a million things that I have not yet had the time, the curiosity, or simply the courage to learn about yet, and these are essentially my resolutions for the new year, that is, to be more courageous, more open-minded, and above all, to listen to people, real people, not just words written on books or voices coming from headphones. I spent my 20s reading ferociously in order to be an interesting person, only to find that I lacked the people skills to share my ideas with anyone. I will not make the same mistake again.