Hello Spring

The trees in the local park are abloom now. It has been raining and the drive-ways on this street are covered by carpets of flowers.

I am getting used to the train. It isn’t without grumbling and mumbling that I have deigned to be one of the commuter proletariat. However, it has its moments of interest. I have started speaking to some of the regulars. The station at the other end is under a bridge, graffitied and decorated by litter of dubious origin, and there are usually more drug transactions than ticket transactions at the early hours I get there. I’ve been living it real these days. I am so glad that I did not move into the city proper. I can only imagine the grit I’d need to survive seeing all this everyday without a break. Now I get to be a sensitive, bleeding-heart snowflake who sees all this on the commute and gets to come back to a bubble of denial at night.

I found a cafe near my new place. It is family-run and they have really good coffee. It is a different crowd from the one I used to see at the cafes near my previous lodgings. There are more entrepreneurial freelancers and small-business owners here. People are more courteous about wishing me good-mornings and hellos. The men, particularly the older generation, are very different from the samples I’ve encountered before in the Valley area. I’ve never been complimented as much as this in my life after moving to California, unless I take into account the intrepid men of Mexico City.

All is not wonderful. They drive big pick-up trucks and RVs, tugging along boats and boards, and eat up narrow streets with little concern for traffic rules or other road-users. I’ve started wearing flashy clothes just to alert them when I am walking along to and from the station. I admit that I only just needed a half-decent excuse to justify wearing flashy clothes.


I need to spend some time plotting what to do after this gig. The more I see the explosion of interest in my core fields, the more I feel abruptly dispossessed as the field is overtaken/hijacked by those from other areas.

There shouldn’t be any gate-keeping. Anyone with interest, passion and aptitude should be welcome, anywhere, in a technical domain. This I am firm about. I am not advocating gate-keeping. We do a lot of gate-keeping forming academic cliques in industry labs, and that is terrible. It leads to insular echo-chambers with little room for borrowing concepts from other fields or bringing in fresh perspectives and solutions. Gatekeeping isn’t good and that isn’t what brings me here to write today.

What brings me here today is that I see new people from other areas inventing poorer solutions to solved problems, just because they have seen it solve different and almost similar problems in different fields. I am left wondering about the state of affairs, about what the slow and painstaking progress in the field has come to.

Robotics isn’t Data Science, but it is getting there. It has come to mean anything and everything, a catch-all for everyone with interest to come and try out things with no interest in looking at what came before. If you sound important enough and carry on loudly and persistently, if you are from a demographic that’s traditionally considered to know more, you are going to make it big right now in this. It is getting to be the Big Data of our times, spinning away from concrete meanings to vaguely important buzzwords. Money attracts a different sort of people than who would do it otherwise. Getting paid makes me happy, of course, and I am glad that we are getting paid what we do, but I cannot help wonder if it has hijacked our field completely.

There is this pervasive mindset that the problems are too hard to be solved without hacks or that they are too easy and can be solved by approximating a single joint distribution from end to end. There is this newfound idea that each new entrant has sufficient knowledge and context to bring in innovative solutions, that hacking and brute force and max-likelihood is how we solve problems, that nothing was solved before because nobody capable was there to solve it. I find people leading teams with tunnel vision, adding chains of if-else, shifting their problem formulations at each iteration to be narrower and narrower as they come to understand that they don’t actually understand, trying to create problems that they’ve got solutions for so that it seems as if they’ve actually innovated, trying to avoid creating and comparing against standard or classical baselines because this is clearly, inescapably superior for reasons. I find people talking about convex optimizations and L2 loss without knowing or worse, not caring, what the objective function should be. I find people reinventing everything from search to approximation, overloading terms, and excusing a lack of logic and generalization because this is just engineering and not science. None of this is necessarily as bad as the fact that we’ve reached a point that looking back at the history of the field is not only rare, but actively discouraged. Everything that came before was so bad that it isn’t worth discussing any of it now. We’ve got to move on from the loser ex.

Every field has its share of talkers. And they are necessary, to get funding, to attract interest, to broadcast advances etc. Some, though, want more say in how things are done, want their ego stroked in a different way by exerting control over the work of other individual contributors. These are people who are attracted by not interest in the field itself, but by other motivations: their day in the limelight, money, proving they are cool etc. These are people with superficial knowledge who posture themselves to be specialists, who are always claiming to know how to do things better, whose verbal output is more than their actual output, bullying, manipulating to get the outcomes they want with little regard to merit or commonsense. What motivates them are often different reasons: ego, getting off on power over another, a fear of being found out as someone who isn’t good enough and has to keep ahead of being outed by posturing etc. In an ideal world, they get ignored and life goes on. In this world, what happens usually is that there are chains of them in the hierarchy, each amplifying the voice of the next, until they poison entire projects as they set policy.  When a field becomes high-value or cool, it brings these  people forward. They aren’t interested in making anything, or selling anything,  as much as they want power and their share of the limelight.  Then they set about excluding who is a threat to what their goals are, overtly or covertly, until they have destroyed an entire ecosystem of innovation.  When the field fails, eventually, they place the blame on the original ecosystem. In the case of robotics, generally the specialists are called out for not being computer scientists, and this is where the blame is laid when projects fail. Robotics is engineering Artificial Intelligence, which is just mathematics and computer science. The Mars Rover didn’t get there without roboticists applying a bit of computer science. Before the era of GPUs and easier abstractions of computing, roboticists were just sitting around in labs optimizing code to exploit architecture and algorithms to do expensive operations for vision and motion. Scaling is a different problem, altogether, but it is perhaps unfair to expect that anyway given the problem formulations of that era.

Assertiveness is what is necessary to combat this toxicity. Assertiveness, unfortunately, is not a part of the standard engineer make in the valley. You see bullying, you see aggressiveness, you see passive-aggressiveness, but you rarely see assertiveness. In fact, many cannot understand the distinctions, being as unused to assertiveness as they often are. And I can see why. We have become creatures that silently suffer injustice until we no longer can, and then we leave with or with out exploding in anger and misery. It is easy for us to jump jobs. So why not? Assertively calling out dark patterns of behavior and decision-making isn’t easy, feels like a losing game anyway, and it doesn’t seem as if it ever changes anything.

In robotics, in the industry, we’ve now so many of these toxic chains of hierarchy, of people who talk and posture and are busy trying to reap in the limelight and the power while it lasts, while pushing back research and innovation by years, squandering away the golden opportunities and faith that investors and others have given us now to make products that are actually tangible, at scale, and in the hands of interested consumers.

Yes, it is easy to impress someone without the background with point-cloud segmentation or A* routing or the output of a CNN, but that doesn’t mean they’ll be fooled for too long. And I wonder if we’ll even have the expertise necessary to explain why the failures exist on the tails when the investors come asking, because we’ve gotten used to hacking together demo after demo to bring in the funding, and cornered ourselves into non-modular systems where a poor end component tries to solve everything because the components above haven’t been designed to be robust, or even designed at all.

Our promises of our systems are often only as truthful as fake online dating profiles. One day, if you meet them in person, they are going to know, and then it’ll be too late to make it work. A sad day that I am comparing a field I used to love with online dating, with a similar number of false positives! I hope it doesn’t kill the field in entirety for another long and sad winter.

I am passionate about the field still. I have enjoyed working with and learning from each high-school student group that I coach over the summer for projects. I have enjoyed mentoring and learning from each intern, researcher or engineer I’ve hired from within or outside the field, each discussion I have with people who care about these problems and want to solve them. I have been fortunate to enjoy an excellent rapport with my mentors and colleagues in the field, in both academia and the industry, and we cherish our debates on the tradeoffs of compute versus coverage. I want the interested or curious outside world to think of us, of our field, as something that is cool, welcoming, and not toxic. However, I am afraid we’ve veered off that track. Perhaps we’ll come back to our senses soon.