If you read my previous blog post, “Becoming an Independent Researcher and getting published in ICLR with spotlight,” you will know how much I struggled to enter the world of machine learning research. However, you will also know my struggles ended on a positive note, “there is light around the corner,” or so I thought. Well, it is complicated. — Vancouver, Canada

9 months after my ICLR spotlight award, as an Independent Researcher.

A real story of having succeeded, to then have promises broken and dreams crushed, and feeling sorry for yourself when others have it worse.

In December 2019, I got published at ICLR 2020 as an Independent Researcher and received a Spotlight award for my publication. The goal, to get enough recognition to become a Research Software Engineer at a known industrial research group.

This publication went about as viral on Twitter as most researchers can hope for. I received many questions and wrote the blog post “Becoming an Independent Researcher and getting published in ICLR with spotlight,” which also went viral. Now 9 months later, more than 3000 new Twitter followers, 2000 emails, and a few “hey’s” on the streets, many of which have asked for an update — well, I think it is time.

My blog post on “Becoming an Independent Researcher and getting published in ICLR with spotlight” that went viral on Twitter.

I think many assumed that everything would be well. That life is somewhat fair, that if I worked hard for a year to achieve something, it would happen. But life isn’t like that. Yes, to address the elephant in the room, it is 2020, the year of the coronavirus. However, I think it is more than that. No matter the reasons, the trauma of having your dreams carelessly dropped by other’s invisible hands, again and again, are the same. But also, not everything is terrible, as you will read at the end, and some have it much worse than me. As such, I really struggle with; what am I allowed to feel when others have it worse?

Many like to think, that success, especially earned success, is followed by more success. But success is followed by reality and reality can be cruel.

In this article, I don’t just want to tell you what happened; I want to give you a sense of how it felt. This is not a success story; I could have framed it that way, that would likely have gotten me more Twitter followers. However, I always want to be honest, and the truth is this year has been emotionally very painful. I hope it will help you, or someone else you know, who may be going through the same, and this will make you feel less alone. — So please bear with me while I explain something fundamental to me: 1) my struggles with pain and 2) my dreams.

What am I allowed to feel?

This question is always on my mind whenever I feel hurt. I know it is silly; there is no law that prohibits what I can feel. However, due to my childhood, this question is engrained in me.

To keep it short, I’ve always suffered from chronic-headache and occasional migraines. During primary school, I had 21% sick days. My teachers’ would insist that “my headaches weren’t real,” that “I just want sympathy,” and that I should “stop making a face that showed pain.” Those last two statements have made me struggle a lot with what I’m allowed to feel.

Today I know my headaches are objectively measurable, I know how to handle it, and I have fewer sick days than the average person. However, those childhood experiences influenced me a lot.

My love for science is a love for truth-seeking. Formed by being constantly asked to question, if my headache were real.

I don’t want to change my past. I value my love for truth-seeking, even though I know there is rarely such a thing. Don’t feel sorry for me. — Greater Copenhagen, Denmark

Why I want to be a Research Software Engineer

I used to think that academia had only pure intents. Listening to my friends outside of research and academia, I think many share that belief. However, once you become a part of academia, that illusion fades, and we see that academia is not an excellent implementation of science.

There are limited funds that researchers fight to get, creating a hostile environment that favors the more cut-throat people. Those people that survive tend to be more interested in power, appearance, and achievements than they are about scientific integrity. Yes, there are truly good researchers out there, I’ve met them, and they are the ones that inspire me.

Simultaneously, I’ve always enjoyed programming. To be honest, this is the only kind of work where I can completely lose track of time. And in my earlier life, I spent a lot of time contributing to Node.js. Today, knowing that my code runs in everything from NetFlix to NASA’s spacesuits makes me very joyful. Even today, much of my income has been from building open-source profiling tools.

Becoming a Research Software Engineer has become a dream. A job where I get to be involved in research and science, through programming, but leave some of the academic parts, that I don’t like, to others.

The destination for me has always been clear to become a Research Software Engineer. How to get there has never been clear. — North Zealand, Denmark

January 2020, so many emails

During January (and some of February), I received nearly 2000 emails from many aspiring researchers. Most either congratulating me, offering me advice or needing a shoulder to cry on. I spent all of January and some of February just answering emails. If you didn’t get an answer from me, I’m so sorry; it was impossible to keep track of everyone.

Among these emails were also offers for Ph.D. positions and occasional job proposals. To be honest, most of these were not relevant. I interviewed with quite a few professors, many of which admitted to me they did hard-filter by publications. When I looked up their own work, I noticed they hadn’t done any top-tier conference publications themself, and they were perhaps looking at me as a way in.

Many professors struggle just as much with getting into top-tier conferences and journals, as young students and researchers do. We just don’t notice them, the same.

The job proposals were mostly startups. There is nothing wrong with that but I’ve been in 3 startups, I don’t really want to do that anymore. I also think it is rare that startups would prioritize the kind of research I like.

DeepMind; hope like never before and the careless hands

Among the many emails, I did get one offering me a shot at achieving my dream. It was from someone above Senior Research Engineer at DeepMind, offering me a recommendation for a Research Software Engineer position at DeepMind. They would personally hand over my CV with their recommendation. The message was, “I should get an introduction call from HR within two weeks, after which they could start interviewing.”

To say I was thrilled, humbled, and joyed, wouldn’t do my feelings for such a gesture justice. It was indeed a shot at achieving my dream. As such, the sadness was only greater when I was never contacted by DeepMind. I reached out several times to my referee, as far as I know, there was no reason for the radio silence, and my referee was disappointed as well. — Keep in mind this was well before COVID-19 had hit Europe.

Your hopes and dreams are held by invisible hands, that can carelessly drop you at any time. Affecting you forever.

Maybe you recall, I had a similar experience with Google in 2019, after my Distill publication, as I now had with DeepMind. Every time I go through this, a part of me is chipped away and a wish to never hope or dream again emerges.

No matter how good you are, the invisible careless hands are just that; careless. Not because you are unqualified. Don’t think any less of yourself because of them.

Google AI Residency; promises, a virus, and tears

I try to always have a plan B. I don’t enjoy it. To have a conversation with someone, expressing my genuine excitement for a job opportunity, while knowing I’m engaged with another party. It feels I’m cheating on the ones who recommend me. But I also know, far too well, how careless the invisible hands at play can be.

While waiting for the DeepMind call that never happened, I asked for recommendations from some of the Googlers that had retweeted my ICLR paper. I had already applied for the Google AI Residency back in December. The application deadline was one day before the ICLR announcements. Therefore I couldn’t say I had been published, and I felt it essential that a potential referee(s) could inform the Residency Committee.

From what I’ve seen, the Google AI Residency is one of the only ways to get a Research position without doing a Ph.D. The Residency program receives thousands of applications but only have a few positions per year, so I knew my chances were small. — Then, one day, I got the magic email. I’m being considered for the Google AI Residency.

The process is; you are assigned a recruiter contact, then there is an introduction call, a programming interview, a machine learning interview, and a final committee approval. I had no doubt I would pass the interviews, but the committee approval is a lottery, and I was trying to mentally prepare myself for the careless hands that I have experienced so many times.

I passed the interviews, as expected. Still prepared for disappointment, I awaited the committee’s approval. However, this time my hopes and dreams were not dropped. I was approved! Even better, I was extraordinarily approved for 4 Research Groups. Apparently, they usually only approve with one Research Group in mind. As an example, a research group would be Google Brain in Toronto.

Due to the extraordinary approval, I had to informally talk to each group to find the best fit. Although a group could technically turn me down, I think everybody considered that unlikely. — I was even told that Geoffrey Hinton was looking forward to talking with me.

A dead naked tree, completely covered in spider webs, somehow consumed by inspects and arachnoids. There are green trees arou
Sometimes nature shows its ugly side, kills the beauty that once was, and there is nothing you can do. — Copenhagen, Denmark.

Then, around April, as the coronavirus had hit Europe and North America. I was informed that the group talks and overall hiring would be put on standby until Google had decided how to prioritize their hiring. I was asked if I could start in January 2021, instead of September 2020, to which I said yes.

I anxiously waited for 4 weeks. I told myself; that surely Google wouldn’t expect to be hit hard by the coronavirus, that being approved for multiple countries (none of which were very affected) would make a good case, or that my recommendations would help. But no, I was not hired, and the committee approval was revoked, I did “no longer fit their business interests at this time.” If I was still interested, I would have to apply for next year like everybody else; my existing approval wouldn’t matter.

I didn’t feel angry. After all, Google is a private business; it is their privilege to do as they wish. However, emotionally, I was totally destroyed for several months. Even today, 4 months after, I still get overwhelmed by the pain of that loss. For sure, this has been the worst event in my adult life.

To have ones dreams come true, to feel sure that it would happen, then to have it all taken away by invisible hands. It makes you wish to never dream again.

You might think: well, I could just apply for next year. If the committee found me qualified once, they will do it again. But these positions aren’t just about qualifications or skills; they are about being lucky, and luck is something I can’t depend on.

Other promises

There were other promises. Promises regarding interviews and jobs, but that all felt apart and not just because of the pandemic. I don’t want to name the smaller companies. Some of the larger were Salesforce and Huawei, just to name a few.

For my freelance business, I also had some project agreements coming up, those too were canceled. So, financially it also didn’t look great.

For a time, I collaborated with a researcher from DeepMind, out of interest, and to keep myself from an emotional abyss. But they too stopped answering emails and went radio silent.

There is much beauty in life, but it is also very fragile. Anything can be washed away, by nature or mankind, at any time. — Stevns Klint, Denmark

Mila; becoming a Ph.D. student

By now, you likely understand that becoming a Ph.D. student was not my first, second, or third choice. However, after so many broken promises, it became the last option that could move me towards becoming a Research Software Engineer.

As the title says, I’m now a Ph.D. student at Mila. It is a great institute, in many ways, and I should be very grateful for having gotten this opportunity. Indeed, there are those who could only dream of being in my situation. However, I don’t feel as grateful as I should be.

This conflict of “I’m not as grateful as I should be”, is something I struggle with daily.

I’m not feeling particularly grateful because I don’t really want a Ph.D. or to be part of academia. I know I’m qualified for the positions I want, I passed the interviews, I have been recommended plenty of times, I have the publications, I was promised the jobs. But somehow, by circumstances, by nature, or by the carelessness of invisible hands, I didn’t get those jobs. Now I’m spending 4 years of my life in an environment I don’t really like, just to become overqualified.

Adding to that, I’m not allowed to travel to Canada due to COVID-19. Instead, I work from home in Denmark, trying to be productive. You might think: well, many are working from home now. But please consider that I was already working from home since March 2019, as an independent researcher. So after 18 months in my one-room apartment, I’m long past my limit. The finances also don’t make any sense, as the stipend is scaled for Montreal, not Copenhagen which is far more expensive.

Post-publish edit: Not everybody made it to the Q & A and therefore end up think I hate doing the Ph.D. That is not the case! As I write later, “I think both Mila and my supervisors are on the better side of the academic experience that I have otherwise learned to expect”. Adding to that, I really can’t give a fair opinion about Mila, as I can’t travel to Montreal and study like I’m supposed to.

Success does not always follow success

I want to get back to my original statement:

Many like to think, that success, especially earned success, is followed by more success. But success is followed by reality and reality can be cruel.

Hopefully, you understand why I say this now. I think, in particular, it is crucial to anticipate some cruelty. If I had assumed that DeepMind, Google, or one of the other companies I didn’t mention by name, kept their promise. Then I wouldn’t have made the plan B, C, D, E, and F; and I wouldn’t be at Mila, which is definitely far better for me than the alternatives, such as doing trivial statistics or basic programming.

There is a thin line, but an important one, between anticipating cruelty and being cynical. My advice is to not expect cruelty from any single person, just anticipate that there is a systemic issue. Additionally, don’t prevent yourself from applying for a position or not asking for a recommendation, just because you fear rejection.

Dealing with loss, when others have it worse

It’s an obvious truth that COVID-19 has caused a lot of deaths. Even without COVID-19 true loss is inevitable. In the past, I have lost people too. I know that in those dreadful moments, concerns like career and opportunities are not on anyone’s mind. So who am I to feel sorry for myself?

If there is one thing I have learned, from living a life with chronic headache, migraines, and a few cluster-headaches, it is that; comparing pain is completely meaningless. What we feel is relative to what we have experienced, learned to deal with, and what is important to us in the moment.

So if you find yourself feeling sorry when others have it worse, know that you have my permission to do so. Use that permission to talk to your support network, use them to validate and process your feelings.

I know, having been an independent researcher, that you don’t always have someone who has been through the same. One of my friends compared me “losing the Google AI Residency” to her having been without a job for a couple of weeks after finishing her Master’s degree. Obviously, it is not the same, and I felt disappointed. However, I remind myself that given her experience of occupational loss, this may be the worst it has ever been, and therefore her pain may be just as impactful as mine.

A palace and grass field. The sun is behind a cloud so everything poorly lit.
Right now, the sun doesn’t shine on us, let’s not lose hope. — The Hermitage Palace, Denmark.

Q & A

  • Q: Do you plan on finishing your Ph.D.? A: Well, I’m giving it an honest chance. I think both Mila and my supervisors are on the better side of the academic experience that I have otherwise learned to expect. However, the financial situation provides an upper limit on how long I’m willing to continue.
  • Q: Do you still think publications are necessary to get into a Ph.D. program? A: Your best chance is to do an MSc+PhD program; those are much easier to get into. If you, like me, were outside academia for at least 2 years and already had an MSc, then I think getting into a Ph.D. program without a publication is still very hard.
  • Q: What unexpected challenges are there in getting a Ph.D., after having been in the industry? A: Recommendations! Ph.D. programs don’t care about industry recommendations. Even if your new supervisor approves you, professors’ recommendations are still bureaucratically necessary. — After 3 years, it is tough to get more than 1–2 recommendation letters, and many programs require 3–4 recommendations. I had to camp outside a professor’s office for an entire day to get one of them.
  • Q: I’m from the future. What is this coronavirus? A: Hi, good to hear it passed. It was a pandemic, look up COVID-19.

Ph.D. Student at Mila, researching interpretability for Machine Learning because society needs it. Published in Distill, SEDL|NeurIPS, and ICLR (spotlight).