Returning to Academia

Despite a great deal of deliberation, I struggled to articulate my thoughts on the subject of “academia” versus “industry” for years. A decade, to be precise. The many debates I’ve witnessed during that time that argued over the merits of one path versus the other always struck me as incomplete, self-serving, or confusing. Until earlier this summer it suddenly dawned on me what, in my mind at least, captures perfectly all that the juxtaposition of the two conveys: Openness.

I’ve come to find that single word to make all the difference in the world. It doesn’t just shape how you think as a scientist and what you do research on, it forms who you are as a person. Its presence brings with it humility, veracity, and diffidence. Its absence, egocentricity, mendacity, and grandiose delusions. Ideologically, one path endorses it, the other opposes it.

The conception of that perspective snowballed into my leaving industry and my being put on a path to transition back to academia. I wrote this note only to organize my own thoughts, but have decided since to share this opinion publicly, albeit an edited and less verbose version of it, so that it reads less like a stream of consciousness. But before I begin—and lest you, the dear reader, should jump to any unintended conclusions—I’d like to state clearly whom I’m writing this for, what my objective is, and how my statements are to be understood.

Target audience I hold a PhD in Computer Science and gravitate heavily towards the “philosophiae” part of my title in its original sense: Love of wisdom. I strive to be intellectually honest and do not trust results without verification. I’m always itching to write to explain my thought process and assess its logical validity, and prefer to read your arguments in great detail for similar reasons. I am self-critical and enjoy being critiqued and even proven wrong. Those are the qualities and qualifications of the kind of person I’m speaking to in this post.

You may be a creator; a scientist who likes to monetize your ideas; an applied researcher who enjoys solving highly practical problems. You will find much to disagree with. That’s fine. But bear in mind that this note isn’t meant for you.

Objective My goal is not to start a grand debate. Nor is it to encourage you one way or the other as you form your opinion and choose a path that’s right for you. On the contrary, I wish to present to you a perspective, so you can think about it and analyze and critique it and contrast it with other opinions, and ultimately prepare yourself for whichever path you choose, or do things differently if you find yourself in a position to effect change.

Disclaimer I’m not naïve enough to paint any institute with a broad brush or declare that there is a total order governing this debate. Nor would I ever consider a claim to be rigidly true or false. No, none of the statements I make are entirely true, nor completely false. Instead, they reflect a Bayesian view: Your experience taking either path is conditional on your prior belief, and the events you experience continue to inform the prior of others. My characterization of industry and academia is only that: an informed prior. That means, there will be—no, there must be—counter-examples to every opinion I hold.

It goes without saying that I continue to grow. My opinions will evolve too. That prior will be molded accordingly over and over again. Will I hold the same opinions a decade from now? I hope not, or else I have stopped growing.

I conclude this disclaimer by adding that, my opinions are not to be interpreted as statements on the labs I’ve worked at, companies I’ve worked for, or people I’ve worked with. My observations are general, and reflect my take on what I consider to be culture. I wish they were specific to a single workplace, because then I could, perhaps, help improve it.

The Role of Openness

This section presents my main thesis on the importance of openness to me and why I see one side endorsing it and the other opposing it.

Open minds

Here is my philosophy: What you work on should surprise you. How you approach it should not be on ideological grounds. You should accept the perpetual suboptimality and eventual death of (your) ideas, and let them reincarnate.

These perspectives, in my experience, are antithetical to what applied research labs in industry stand for. You are seldom allowed to explore into the dark and discover parts of the scientific literature that are foreign to you or are tangential to your mission. Instead, you are pigeonholed into one role and are expected to stick to it. Exploit, in other words, rather than explore!

Letting your mind wander into areas unknown to you is a necessary step in gaining breadth, forming a vision for your research, and making serendipitous discoveries. That same act is also imperative in developing solutions to research problems.

Alas, how you approach problems, too, is often ideological. You face contrived constraints imposed on you by arbitrary engineering decisions, product requirements, or worse some powerful person’s opinion; your exploration of the trade-off landscape between various algorithmic approaches is not steered by facts, it’s determined by what identity your institution has decided to adopt.

Oddly, that makes sense. Businesses, including, contrary to popular belief, most start-ups and their investors, are risk-averse. There is little to no incentive to innovate. Recycling as much as possible is preferred over reinventing. What makes less sense is the disinclination to let go of dead or dying ideas. Bad ideas are kept alive on life support, not re-evaluated, tossed out, or reworked. Sensible or not, inertia is ever-present.

In academic research institutes, on the other hand, you become irrelevant if you prioritize exploitation, ideology, and stale ideas. It is by remaining open to the idea that your work is already irrelevant the moment it is published, that you motivate yourself to explore uncharted territories. It is by seeing the flaws of your own arguments that you continue to innovate. Importantly, you determine the state-of-the-art objectively, not by hiding or bending facts and elevating your ideology.

Having an open mind is essential in conducting long-term, innovative research. That simply is incompatible with industry principles.

Open debates

Another part of my philosophy is this: Separate your ideas from your self. Criticism of your idea is not criticism of you. Learn to think critically, and to accept harsh but objective critique of your work.

Having to publish your work and being measured by the volume of publications and number of citations have many downsides. But the one fantastic opportunity that having to put your ideas out there creates is that, it allows others to judge your work and critique it. That is what academic research is all about: Critical examination of scientific works.

That creates a culture where ideas evolve into something far better over time. I can recount many of my own published ideas that have been picked up by others and improved substantially over the years. Even a single journal publication that goes through multiple rounds of reviews becomes more complete after every round. Reviewers hone their critical thinking skills, and authors perfect their ability to receive critical feedback. It’s not easy. But it’s done, often, and always. It’s part of what academic researchers do. It’s not always perfect, but it at least exists.

That culture simply doesn’t exist in industry. Ideas are evaluated monetarily, not scientifically. There is no direct critique of a research idea, only indirect assessment of its potential to improve margins and increase profits.

Even the machinery to critique ideas doesn’t exist. That makes sense. Companies have an incentive to protect their intellectual property, lest more competition emerges. Yes, you can publish your work in academic forums, but that’s unimportant to your employer and its outcome carries little weight. Often the published work is so far removed from what actually happens that the publication can in no way represent a test of the strength of the respective company’s product. External scientific feedback on your publications is uninteresting unless it leads to improved monetization opportunities.

Open books

Finally: Security through obscurity does not last. Accountability matters.

Companies and their investors care about their bottom line. To protect that bottom line, their operations are often shrouded in secrecy. How they do things is a tightly-guarded secret. It is only through high-level, vague, hand-wavy language that they describe how they do things differently. Obscurity is a must.

It’s not hard to see how obscurity and lack of scientific accountability creates a breeding ground for flagrant lies and fraud. When definitions and methods of evaluation fail to be open and rigorous, truth can be bent in marketing campaigns.

It is a daily occurrence. Somehow we are all convinced, as a timely example, that there is an “if and only if” clause joining the generation of semi-coherent sentences, and intelligence, and that that’s the magic bullet that will fix everything—and what everyone must fear! “Define intelligence” you may inquire, or “evaluate intelligence” you may ask, and you will get vague statements at best. Never mind that, my dog can’t speak or parse language, but displays sophisticated behaviors, including hints of Theory of Mind. Never mind that, people can speak languages but are less “intelligent” in more ways than one than many other species (e.g., emotional intelligence, processing of sensory information).

There is no risk to hyperbole. But there is in academic research. Your ideas are published and critiqued. Openly so. Diffidence is a virtue. Claiming to “revolutionize” this, “democratize” that, or design “next generation” that, is not (or, rather, should not be) in the vocabulary of a scientist. Grandiose delusions do not stand the test of time, and damage your scientific reputation and standing; something companies need not worry about. They come and go, rebrand as needed, or lay teams off to pretend they are changing.

I find that culture, that is so prone to exaggeration and mischaracterization, to not be conducive to scientific research.

Concluding Remarks

There are many researchers who can’t imagine not working in industry, in applied domains and on practical problems. There are many academics who, I personally think, would enjoy industry and would be more well-placed there instead of in the classroom. This piece is for scientists who care less about applications and more about making (often obscure) discoveries. That’s me. I penned this essay for me.

I wrote this to understand why I’m burnt out, all the time. To pinpoint what I needed to change to do better. I realized through this exercise that openness matters to me, and it is the reason why my greatest accomplishments in my decade in industry have been my academic activities, not the software I helped create, or the billions of user requests my services served. That was it for me.

I now know what my next decade should look like. Will I manage to realize that vision? I’m trying. Will my views change over time? I hope so, I’m open to reshaping my perspective. Will I continue to work with researchers from industry? Absolutely; some of my greatest mentors work in the tech industry. But do I see myself as the greatest fit for industry? Perhaps not.