Bernhard Egger is a postdoc at the Computational Cognitive Science Lab at MIT. His research is focused around inverse graphics in Computer Vision and Computational Cognitive Science. You can find out more about him on his website https://eggerbernhard.ch/ or on Twitter https://twitter.com/VisionBernie.
Hi Bernhard, thanks for joining How I Fail! Next to your official bio, could you say a bit more about yourself, and what made you join this series?
I want to figure out how vision works: how human vision works, and how we can build computer vision systems that generalize. I choose to perform computer vision research based on an inverse rendering approach. I am convinced that Analysis-by-Synthesis is the way human vision works and the way we should build machines that perform vision in such a way. This idea puts me in an interesting niche in the community. During my PhD it sometimes felt like fighting a lonely battle against a huge flood of supervised deep learning research in the community – however over the last few years inverse rendering, 3D Morphable Models (3DMMs) and in general more model based approaches to vision are coming back. Even conferences like CVPR (the most important one in my field) are going more and more in that direction. I might have chosen a path that is not particularly fruitful when it comes to citations, but it definitely enabled a lot of steps in my career. Whilst a career in academia was less a master plan, than a “just in time” evaluation at each step, I would say it mainly worked out because of the niche I’m working in, that I found for myself. This niche for me is fun and therefore I can make an excellent, motivated story for a presentation – which I think is crucial for i.e. (faculty) interviews.
I wanted to join this series mainly with the intention to show that not everything is as smooth as it might look in my CV, a lot of things that are important cornerstones in my CV had a significant component of luck involved – and there is also a lot of failures or drawbacks that are just not mentioned.
What’s a memorable failure for you?
My biggest failure was probably my first attempt to get a personal career funding grant – I was following my partner to Boston, so I cold-called professors at MIT to get a postdoc position (a cold-call is how more than 40% of the postdocs get into MIT). I intended to get my own funding, by applying for a grant for 18 months, with an acceptance rate of about 50%
After visiting a lab at MIT and having support from a professor for the grant I was super pumped and confident that it would work out. I wrote the grant during a ski vacation (spending the days on the slopes). I then got feedback from the professor and also from my peers that it needs some work. I had one week to go – but I was super confident that with MIT as an institution and this high acceptance rate it should work out. So I did a little bit of work, but did not invest as much time as my peers would have recommended.
In the interview (which is short <10 minutes) they basically took me apart and isolated the weaknesses of my proposal. They caught my vague formulations and found that for some essential parts of the proposal, I was just not certain enough how to do it.
I was first disappointed and angry at the reviewers – but then quickly realized that this was my own fault – I took it too lightly and felt like a small kid testing his boundaries of what is acceptable and what not.
When preparing the revision for the grant, I realized how sloppy it was – I spent a whole month writing a proper proposal and got it funded in the second attempt. This however also led to a delay of my start at MIT and I moved to the US with a delay of 3 months after my partner.
Another big failure was an attempt to organize a workshop on 3DMMs at a major conference. We had a group of organizers together and wrote a decent proposal. It was rejected from both CVPR and ECCV (two major conferences in the field) in years where there were tons of workshops on “deep learning for x”. This was very frustrating, especially because a workshop should add content in an area that is not covered enough by the main conference. So in my eyes a deep learning for x workshop might not add much value to a conference that is anyway 80% deep learning for x.
After those two rejections, I basically gave up – but a senior person among the organizers told me “Let’s try to do a Dagstuhl seminar instead” and that worked out and was my best scientific experience so far!
Guess which conference added 3DMMs as an area two years later? CVPR… But it is not good enough as a CVPR workshop 🙂
What kind of topics are associated with failure, and are not talked about enough? What problems does this create for early career researchers?
I think academia just has this huge illusion of a perfect world: perfect CVs, perfect projects. When listening to a presentation, especially from a senior person the story will be so smooth it just sounds like everything was always planned that way. Some people include some failures in their presentation, but then more in a funny way than formulated as a real failure.
I really appreciate that you added a CV of failure to your website – I think it takes a lot of courage. It is refreshingly transparent and it would probably be valuable if more people would do so. However it might come at a cost: Will I be judged by others based on sharing my failures? And especially if there are only a few people sharing their failures it is easy to judge them. I don’t know what the right solution here is – everything public and transparent might be a bit much, but sharing failures within the lab, especially from the most senior person might be valuable and create a culture of sharing failure.
One other key topic that is prominent in academia is the imposter syndrome. I stumbled on an online test recently on twitter (http://impostortest.nickol.as/). And if you check #ImpostorSyndrome you will see tons of results from researchers with high scores in this test. We are in an environment where smart people come together, and almost always you will find somebody appearing smarter and more successful. It is sometimes really hard to not feel stupid in this environment. One important message here is that ~ 70% of people in academia feel that way – we can’t all be that bad 🙂
Is the situation changing? And are there differences between different fields, countries etc?
I have hope that things are changing slowly. The first important step is probably awareness. Some institutions put big efforts into the mental health of their students and employers but some of the issues might be in the structure of how academia is set up at the moment. So fixing the final problem might not be as good as trying to change the cause.
The main responsibility in my eyes lies with the advisors/mentors. I think the moment we mentor someone (starting from an undergrad level, over PhD students and including young faculty), we actively propagate the way we think academia should work to our students. A lot of my behavior and the way I see academia I learned from mentors I valued. If a mentor is particularly bad when it comes to real life challenges like mental health issues that is likely to dramatically affect the career of the mentee and pushes especially minorities out of the field. I also believe some of this can start with small things. I try to make authentic and honest statements that signal that there is room for discussions or self-care and hope this helps to do a small step in the right direction.
Is there anything that you are currently failing at yourself?
Yes – focusing. Over my PhD and postdoc I however started to accept it and just ignore all this self-optimization stuff 🙂 I still see it as kind of failing, mainly because it is easy to spot that I just lost some hours doing X (or nothing) and it feels like there is a constant drain of potential.
Are there specific people who have been influential to how you approach setbacks?
For me it is not really a specific person, but probably each one of my peers. The longer I stay in academia the more I realize that nobody is perfect, everybody gets grants or papers rejected. This definitely helps in the moments where a setback happens to normalize it and look forward. One of the most important strategies that I think I learned from others is to always look ahead – and usually some of the feedback of a review process is useful to improve the work to submit it again. And no, reviewer 2, I don’t think that I should just do deep learning instead!
Are failures something you discuss with the people you supervise/mentor?
I realized that failures like rejected grants or paper annoy me a lot, especially at the moment of the rejection. However after a few days I’m over it and either resubmit it or forget about it. I’m however terrible at giving good advice in discussing failure or rejections. When submitting a paper I already know that there is a big chance of getting a rejection, especially when aiming at the top venues – so once the rejection is in the inbox I’m not really surprised, usually annoyed about at least one reviewer, but I’m just bad in finding the right words to express my thoughts to the first author(s). I try of course to motivate them, and tell them that this just belongs to the experience in academia but what do you say the second or third time – especially if you are convinced that the paper is actually solid?
What types of things do you track regularly?
I’m crazy about my google scholar profile – I check it way too often and am always super curious who is citing which work. This was especially frustrating earlier in my career where it took super long till something changed and some citations even wrote something wrong when citing the work in their related works section. But recently it changed from being frustrating to be motivating – mainly caused by a nice survey paper that gets lots of citations!
That sounds familiar! 🙂
What about successes, should we be sharing more non-traditional ones, not like jobs or papers, but what you are proud of today?
I think such things can be very valuable, especially in a lab setting. Share success or setbacks with the lab in regular meetings. Here I really like the idea of a common coffee break which is common in Europe and not so common in the US. Such a coffee break is usually the point where smaller or different successes and smaller setbacks are shared and discussed and this interaction is definitely highly valuable since it is much more granular and also more personal than a paper acceptance or rejection. If somebody tried a new recipe at home and thinks everybody in the lab should know it – the coffee break is exactly the right platform. Or if you were waiting in line to submit a certain bureaucratic form for 30 minutes just to learn that you also need another one – that is another thing I think can find space in the coffee break. And there will still be enough space for scientific discussions or programming questions even with these little bits of personal success or drawbacks.
What is such a non-traditional, recent success for you?
I ran a marathon during my PhD and I was not shy about sharing it with everybody 🙂 The best part of it however was the training. It was together with a postdoc friend in the same lab and we did all the long runs on the weekends together. I really enjoyed all the conversations, especially since the brain seems to work really well when my body is busy with some other work 🙂 The final moment of then running a marathon and being able to actually finish it was a moment where I was proud of myself and just enjoyed that I actually did something that felt impossible half a year before (I could not run through a half marathon before training and it is not that I didn’t try).
If you’ve been a reader of How I Fail last season, what are your favorite lessons from the series?
I really enjoyed learning about the posters centered around mental health from Dr. Zoë Ayres:
This is an excellent and valuable resource that I plan to share with all my incoming future students since they bring a lot of things on the point that people rarely talk about.
What would a ten year younger you think of you now? What advice would you give to that person?
I tried plenty of those self-optimization and productivity tools. I would tell myself to just not care about all of that and not stress out about (productive) procrastination.