Rohit Krishnan on bureaucracy, experimentation and problems of scale
Uri: I’m excited to be here today with Rohit Krishnan of Strange Loop Canon, who has burst out the gates this year as one of the most interesting and prolific bloggers in an indescribable corner of the internet that I vaguely describe as “tech culture writing”.
Rohit, something I’ve been interested in from reading your writing is whether there’s a few core lenses you could articulate that underlie your worldview. I don’t know if you read Matt Levine, but one thing I always enjoy about him is that I think he brings some clear and consistent lenses to bear on a lot of different situations, and every edition of his newsletter is like “here's some applications of my classic lenses to some new situations.” He’s given a bunch of his lenses names, like everything is securities fraud, and I find that very helpful too.
Your writing is truly prolific and wide-ranging – your recent blogposts range over writing science fiction as a high-impact career, why we should construct more places that evoke feelings of awe, and a piece I can’t summarise about mechanisation – but I’m not sure I’ve been able to distill the core lenses from your work yet. So what I want to know is, Rohit, what is the Strange Loop Canon worldview?
Rohit: That’s a very good point. I don’t think my lenses are nearly as clear – hopefully by the time I write for a few years, it does end up clearer. But I’ll give you a few nuggets that I feel I have written about quite a lot, whether true or not, and have ended up becoming part of that worldview.
“Amongst the trilemma of incompetence, malice, or bureaucracy, we are all too quick to pick incompetence or malice as opposed to understanding bureaucracy”
One is the fact that I feel we do not do nearly enough justice to our existing institutions. I feel we are very good at criticizing when it doesn’t do something that we want, but we don’t ask ourselves nearly enough as to why that is the case. Amongst the trilemma of incompetence, malice, or bureaucracy, we are all too quick to pick incompetence or malice as opposed to understanding bureaucracy.
Bureaucracy here is Weberian, a set of efficient and rational rules and processes to de-bias and get things done. This is one of the themes that I feel I’ve written about quite a bit. We add rules to a system to make it fairer and more tangible and address various inequities and end up with a complex web of rules with unclear boundary zones and interfaces. And some of that is a sympathy for having seen large organizations work internally. I’m not saying you can’t have an organization without the dysfunction, but without a little bit of empathy for the dysfunction, you’re not going to be able to fix it. I think that’s one of the things that I come back to quite a lot.
When you throw people into an organization and have any form of information management, goals, oversight and control amongst them, you’re going to have stupid things that happen. The dumb outcomes are when the various rules interact, and like two medicines that contraindicate. That is normal, right? I mean having stupid things happen is not a sign that institutions are not functioning normally. It is a sign that institutions are functioning normally and you need to kind of work within those parameters to make them work better. I think on an overarching basis, that’s one of the things that I think quite a lot.
“We underdo experimentation generally across all levels for all things, almost all of the time”
One of the other things that I feel I write about quite a bit is we underdo experimentation generally, across all levels for all things, almost all of the time. We as a society don’t experiment a lot even in areas where we should be experimenting a lot more. I say this as a venture capitalist who is ostensibly interested in investing in highly experimental, innovative, interesting companies, but who likes the ease and comfort that comes with trying to find what others find valuable and copying that, much like the rest of the industry.
But despite that, at the object level, whether it’s in science, physics, chemistry, math, etc. or whether it’s in talent acquisition, or management funnel, or whether it’s in teaching, whether it’s in child-rearing, whether it’s in the different ways of organising structures, we do not intentionally experiment much if at all. And worse, we don’t take advantage of the unintentional natural experiments to gather data and learn what works best. I feel that one of the longstanding frustrations in terms of the sort of the stuff that I’ve written, again, comes back to that problem again and again.
“We really do not understand problems of scale”
A third one, if I can throw one more log on the fire, is I think we really do not understand problems of scale. I think in many ways, we look at the world of institutions and super structures the way we look at King Kong or Godzilla. You look at that and go, “Yeah, yup. I mean of course, it’s a giant reptile that stands 100 meters high and this is how it works.”
But the interesting thing about these movies is that they are complete fiction, right? You can’t have an ape that is standing 200 meters tall or whatever, 150 meters tall, I forget exactly the height, because it’s physically impossible.
And I think we need to have the same understanding about scale for other things in life as well, including in building large organizations. You have an internal friction in terms of how large you can credibly make it before it starts breaking down for a whole host of internal reasons, like internal ossification or information transmission getting much harder, etc.
It’s something that I come back to often in my essays. I think as a consequence I feel like I have become a little bit of big government apologist, even though that’s not really my political predilection just because I’m saying “if you want to do all of these things for a country of 70 million in the UK, or 300 million people in the US, guess what? That’s what this looks like.” You can’t really do something at a massive scale while also expecting a highly nimble, easily responsive team. Those two don’t quite jive together.
So these are a few of the lenses that I suppose do pop out. They’re mainly linked to the tech-culture-nexus and to the idea I’ve been toying with regarding the rise and fall of modern institutions. And if I can remember more of the stuff that I’ve written, I could probably talk more about it.
Uri: Phenomenal. So, can I ask you for examples?
Rohit: An example, sure. So, for the Incompetence/Malice/Bureaucracy trilemma, an example is when you want to get a mortgage from your bank, most of the time you have to go through several barriers and departments, each with its own form to fill. It’s the type of thing where each particular rule makes sense (AML rules, KYC requirements, Chinese walls amongst departments), but the final product is basically unusable.
Another example here is Scott's recent article about the overuse by journalists of "no evidence" regarding health outcomes. It's because they have multiple incentives - eg to be factually accurate, to hedge their bets, and to make sure the public interest is served by not seeming like they're underplaying risks, You don’t need incompetence or malice to explain it, even if people on Twitter keep alleging that.
For experimentation, one salient example is that better covid vaccine regimes and restriction regimes would've been useful upfront! Right now all the restrictions are individual and the virus emergence is from collective action, a major mismatch. Moreover, we have drawn barely any lessons from the accidental experimentation that’s gone on, whether that’s Sweden vs Taiwan vs UK or Florida vs California.
Regulations are another great one, especially on housing, where there's almost no experimentation anywhere. You could learn from the different housing policies in SF vs New York vs London, but even that low hanging fruit isn’t taken up. You can take, for example, the conversation you had with Lars Doucet on Georgism, on land value taxes. It sounds like a great idea, but I’m worried about the unknown unknowns which we can only see once we try it out somewhere. So why don’t we!
At the same time, UBI experiments here have been a bright light! Though still the actual data collected or analysed is so little, there are actual policy experiments that have actually been run.
For Problems of scale, I think it’s perhaps easiest to see in large companies:
- Google’s ability to innovate is hampered by their scale
- Facebook’s ability to respond quickly to incidents is hampered by their scale
- In general, ability to respond to competitive threats are hampered by scale, especially once powerful and entrenched
Government programs are an interesting case, because they combine all three. There are often layers of bureaucracy, with a soupçon of incompetence, a complete lack of experimentation, and the problem of massive scale.
Uri: Amazing. Well, shall we talk about one specific example in more depth? Bonus points if you can somehow show the approaches of each of your three lenses to it.
Rohit: I think the perfect example is the FDA – I think it’s what gets the most attention these days, right?
So I think it's very easy to look at that and make high-level pronouncements that the FDA shouldn’t exist, or they should be completely revamped, or their incentives are all wrong. These are very highly specific issues that kind of hint at the broader problem.
Let me walk you through all three worldviews in terms of how that might actually come about.
The FDA: Bureaucracy
First worldview, you look at what has happened with their approval processes, and the bureaucracy that actually exists in terms of getting something through the FDA. Look, the normal scenario for when they approve vaccines and new treatments is like three years, four years. And here, for Covid, they’ve done it in 6 months, 8 months. So you’ve got to have some appreciation for the fact that they have shown the ability to actually move faster when the world called on them to move faster.
And there’s a set of reasons why they haven’t been able to move even faster. Most of them are silly reasons and we should absolutely work to change them. But understanding them is a great place to start.
For example, the FDA does need to inspect facilities that actually manufacture a bunch of stuff in terms of the drug. So they can say that XYZ drug is actually safe and it’s efficient, but they also need to figure out whether you’re able to make them to spec according to whatever quality is acceptable. For Covid, for instance, the way they got EUA was by not doing facility inspections beforehand that got them in some hot water from everyone.
The thing with bureaucracy is that setting up a meeting actually takes a period of time because there’s a set of information that needs to be vetted by a bunch of folks and go through proper channels. But blaming the agency doesn’t actually help change that. So is it incompetence that the FDA wasn’t able to approve the Pfizer vaccine or Moderna vaccine and make them fully qualified earlier? Is it malice? Or is the fact that it’s a rather large bureaucracy, where even when this kind of stuff is working as intended and at warp speed, it actually takes longer? I think it’s very much the latter.
So to me, this is an area where I actually end up slightly sympathetic to the FDA, which is admittedly a very odd position to end up in. I kind of looked at them as like, look, you have an organization that is 15,000 plus people, you’re tasked with basically looking at every new drug imaginable. Not just that, you’re also looking at all of the medical devices and a whole bunch of other unrelated cruft like cosmetics, pet meds, blah, blah. Overlapping rules competing against overlapping regulations until the web blocks out the sun. So their job scope has effectively been blown entirely out of proportion.
And when you’re relying on multiple sets of rules to guide your actions on everything from cosmetics to vaccines to asthma inhalers to smartwatches, then your ability to move is hampered by the rules you’ve created. The political oversight they have at all senior levels, because we like transparency remember, also makes them prone to failures of nerve. None of us like having too many strictures on our actions, but even in organisations the consequence is the slow pace.
The FDA: Experimentation
The second thing, in terms of experimentation, is one where to me part of our frustration with FDA is that they’re a single line of defense. I mean it’s like we have created a ginormous castle or bulwark against incoming frauds. And we look at the FDA and we say, “OK, you are the one standing in front, go look at every drug that is coming in and actually just come up with some rules – use whatever resources in order to say yes or no to all of the stuff that’s coming up. Just don’t let thalidomide in.”
If people are experimental re something like an FDA, what we would do is just set up a bunch of different types of agencies, which weirdly enough, we already have all around the world, use the bunch of different types of agencies, actually assess all of them on how fast they are able to move, how many meetings they are able to take, how many things they are actually able to approve, what is the false negative/false positive rate in terms of stuff that they have actually been able to approve. Create usage exemptions for some of the tools for some people who actually want to kind of try it out just so we can actually see what’s going on.
We need a far better data-gathering apparatus here. Maybe the CIA can help. In the US, 50 states have 50 different regulatory regimes. I don’t understand why there are not 50 different experiments going on at any different point in time to analyse these kinds of things. And people get worried about the word experiment but we need them – the FDA as it is today is the perfect answer for experiments. We need to have, I don’t know, 50 FDAs, 100 FDAs, such that you can actually test their efficacy and evolve. So if the 50 mini FDAs are competing with each other, you actually have a way to change the top FDA without being worried that any change you make will actually result in yet another crazy catastrophe like Fen-Phen or Rezulin.
I wrote an essay on this – called Meditations On Regulations where I asked, “Who guards the guardians?” And part of the reason for asking that is the second question, because there is almost no experimentation. There is very limited ability to evolve, like in laws you have sunset provisions, that are underused. We need to try that a little bit more, just to encourage a lot more mutation and selection within the idea space.
The FDA: Scale
And the third thing – it’s probably the easiest one to explain – is the problem of scale. The FDA has 15,000 people. If the exam question in front of you is “you want an organization to move really, really fast on areas of high importance like during a pandemic, you want them to be able to approve something,” expecting that organization to be the same one that has been trained to go overboard in terms of safety, with that scale, is almost impossible. It’s like expecting Godzilla to run fast. It just wouldn’t happen. I think they are not designed for that. You’re trying to force an organization designed in one way to work in another way.
It’s like asking Google to create the next big social media network. They’re not designed for it. So of course they’re not going to be able to come out and do it in any sensible fashion. There is a reason startups are regularly able to build things and now outcompete the larger companies. It’s not just because the startups are better. It’s because if you have a thousand teams of more people creating new ideas, some of those new ideas will outcompete the 15,000 people organization. But the only way to do it is you need to fight the problem of scale with something that is likely more nimble.
I mean, think about an organization of 15,000 people, even if everybody agrees on what needs to be done, for information to trickle up from some sub-committee to the top people who actually get in a room and meet to say yes or no to something, I don’t see how they can actually do it faster, not under the current strictures anyways. So to me, this is a great example of where the third principle applies.
I mean now that I have said it, I’m kind of thinking whether the FDA was the right one to pick. Many people will dislike it. But I think it’s a good object example that we need to have a little bit of humility in trying to understand the bad stuff, the stuff that actually causes delays within these organizations, is quite often not the result of them being uncaring, malicious actors who are like, “Yeah, yeah, I mean another 50,000 will die but I don’t care. I’m going to take my 4-day vacation before coming in from the long weekend.” If that’s our assumption before going into any kind of conversation then obviously you’re not going to convert anybody. I think that kind of aggressive stance is just genuinely bad for any good faith argument.
Now, granted, it is kind of true if that’s the argument you are having within our Parliament or the Senate or whatever where, rightfully or not, some expectation of malice actually gets you actually a long way towards understanding the true nature of things. But in most things in life, I feel these things actually work pretty well.
I think some of it is also informed by the fact that I grew up in India, right? So for my first 17 years that I was there, I was born and brought up originally in Kerala which is one of the few places where communist governments are actually regularly re-elected. I know what really bad governance looks like. This is not it. I mean this is – I mean it’s not great, and it's bad on some dimensions and it might get worse, but the rhetoric does not match the result is all I’m saying. We have a long way to go, we can fix these things, there are so many levers we can pull, that I feel annoyed that we don’t pull them. Instead, we kind of say, “Oh, shit! The best is behind us. We are all screwed and we are all going to die.”
I remember when I got a scholarship to go study in Singapore, and I remember the nightmare that we had to go through when I was like 17, 16 to get my first passport. It was multiple office visits and asking for bribes in order to do some rather basic paperwork. That’s malice. It is absurd what folks have to actually go through in the world.
I mean, I’m not a fan of British paperwork that I have to do here either but my God, man, I mean our institutions actually still do work, which is kind of insane when you think about the rhetoric. So that’s why I feel I have become a little bit of an institutional apologist, if I can call it that, because if you don’t understand that they are actually okay, quite often, all that’s going to happen is that we replace today’s institutions with more institutions which look exactly the same. Because if the problem is incompetence, you can replace the people, right? If the problem is malice, you can replace the people. But if the problem is bureaucracy, that’s structural.
You can’t just replace the people. You have to change the incentive system. So that’s why to me, understanding institutions is so critical.
Uri: Wow. I think you managed to tie it all up, the three Strange Loop lenses into one neat example.
Rohit: I’m as surprised as you are.
Uri: Ahaha, I have to be honest, I’m not surprised at all.
I should really let you go, but I can’t resist one more question. Which is, can you describe any obvious cases where your worldview has different predictions or implications that the worldviews of other people you see around you?
Rohit: I can try. So I see a few key differences between these lenses and what I usually see used elsewhere.
One major difference from my lenses is that it makes me immediately more sympathetic towards the major fuckups in the world than would otherwise be the case, which has come up a few times already. I think, through these lenses, my “base rate” is adjusted towards a lot more empathy. In Ender’s Game, Ender says “I think it's impossible to really understand somebody, what they want, what they believe, and not love them the way they love themselves,” which definitely seems somewhat true.
A second major difference is that my lenses in general are far less downside risk-sensitive than the general public opinion. Partially this is because if you’re ok with more experimentation, by definition you’re okay trying to chance a negative outcome or two in order to learn more about the world. This holds true for both things like challenge trials for Covid and for things like wealth taxes.
A third is that I feel I'm way more optimistic about these things in the short run and cautious in the long run. So I think institutions such as we currently have work decently well, and also that they inevitably get ossified. This means our belief in institutions shouldn't be like, say, constructing the Colosseum, able to stand for millennia, but rather more like the Ise Jingu shrine where it gets rebuilt and buffed out every couple decades!
What this means is I'm perhaps a little more meta-contrarian in my affect. I worry less about how we live in an anti-Panglossian world, the worst we can imagine, but rather in a more chaotic world where we can build our own meaning and tinkering without knowing the shape of the entire thing is our destiny.
Uri: What an excellent place to end things. Rohit, it’s been an absolute pleasure – where can people find your work online?
Rohit: This was so much fun! I’m on www.strangeloopcanon.com and Twitter. Cheers Uri!