Lazy tech analogies

Chris Gates:
I mean, pretty informal, but I re-listened to the episode we did last week about marginal stupidity. I was talking to my partner about it, and she was like, "So Danny's making a case for liberal arts education?"

Danny Crichton:
Yes, that's true.

Chris Gates:
Is that true?

Danny Crichton:
Yes. Yes. Well, I mean, this is the argument for critical thinking, liberal arts education, for the humanities for consilience, which is you want to have this ability to bring in all these different types of knowledge streams, all these different forms of knowledge, be able to synthesize it together. And you don't synthesize in science, and you don't really synthesize in engineering. In the sciences and engineering, in STEM, you have skills, and use those skills to produce either new research or produce a piece of software or a new chip.

And you have to think critically within that little domain and context, everything about the molecular structure of silicon in order to produce a chip. But understanding the broader world, the economy, how the world's going to function, how businesses, organizations, politics, all these things come together, you have to have a much wider aperture to understand that. And that's, I think, exactly where the humanities come in.

Chris Gates:
I was like, "Oh my God, Danny actually validated my education." I went and did a bunch of things that seemed like it was worthless, and it fed into my job every day, actually.

Danny Crichton:
Well, and just like the classic Steve Jobs anecdote about taking topography at Reed College, which was like, "Well, why do you need to take topography?" And then you actually built the Mac and you were building fonts, and suddenly you were like, "Well, we're going to have amazing topography in the Mac," which is still true to this day, in 2022 and OS 10.

Chris Gates:
Moving on to a completely different subject. So we just did this event with Roger Perlmutter, who runs Eikon, which is one of the companies we are invested in. He gave a really big talk, very inspirational, but there was this one piece that stood out to me that I wanted to ask you about, and maybe you have an idea of it, and maybe you don't. I just wanted to see. But Josh and Roger said that, "We actually have no idea how aspirin works," and that was brand new to me. Did you know that? Do you have any insight into that?

Danny Crichton:
I don't know if I can say anything specifically about aspirin. I can tell you that we don't understand how most medicines work. So-

Chris Gates:
Really?

Danny Crichton:
... that's probably a much more generalized comment. I mean, the body is extraordinarily complex. I think every decade in the modern... last couple decades, we've always had this belief that we're just this close to understanding everything going on. We have the human genome project, and then we're going to finish that. I think we even did an episode on Securities about that of, "Okay, we've sequenced the genome, now we're going to know everything." And then we were like, "Oh shit, there's epigenetics. Genes regulate genes."

And suddenly you were like, "Well, there's a whole other layer to understand what's going on here that we're just..." Just in a recent newsletter, where I was just talking about epigenetic therapy, instead of genetic therapy. Gene therapy has been in the discussion for what, 20, 30, 40 years. Well, now it's epigenetic therapy, which is to say, "Well, what if we don't change the genes themselves with CRISPR, but we just exchanged the expression of those genes in a certain way, so that it doesn't get expressed, even though we're not cutting it out."

Now, scientists are starting to understand that at the epigenetic layer, epigenes actually affect epigenes, if you will. And all biological systems, there's feedback loops. So as you express one gene, it affects the expression of other genes, and it's part of a system. So when we look at drugs, oftentimes, we don't have any clue the mechanism of how it actually functions. We oftentimes can figure out the pharmacodynamics and pharmacokinetics, which means how the drug gets processed in the body. So oftentimes it's processed by, let's say, the liver, and therefore, we know if you pop this pill, it'll last about 24 hours. And then it'll be kind of expunged out of your system, and you're going to pee it out, basically.

But outside of the graph, we don't really understand the mechanism, and particularly in areas, and I'm thinking particularly in neurology. Think most psychiatric drugs, most mental health related pharmaceuticals, which obviously have been in the news a lot in the last couple of weeks and months, we have no idea how they work. None. None.

I remember an interview from the former chair of psychiatry at Stanford Hospital who was like, "A lot of these drugs in a hundred years from now will be known as the bloodletting."

Chris Gates:
Right. The leeches, just like-

Danny Crichton:
It was like, "Because that's how little we actually know how any of this stuff works. We're just throwing things at your brain, and we kind of vaguely have a sense that this helps or this helps." And that actually descends back to very basic drugs. We don't have a model for pain. We don't know why pain comes unless it's an acute, like, you have a nerve pinched in your spine, and therefore, we know there's this pinch. The nerve is getting cut off, and if we unpinch it, the pain goes away. That's a pretty clear model of how the pain works.
But a lot of pain is not like that. If you have any kind of focal headaches, if you have a migraine, we have no idea what a migraine is. Look, I love wine, as you know.

Chris Gates:
Yes.

Danny Crichton:
And so every once in a while I get a hangover, randomly, and I'm like, "Well, why did I get a hangover? I'm going to look on the internet. 2000 years of drinking, thousands of years of drinking alcohol. We must know how ethyl alcohol affects the brain." We got nothing. We have not a slightest clue that... It's like the most you'll get is maybe if you drink some water. It's kind of dehydration correlated, but it's not. You can drink a lot of water and still have a hangover. We have no idea what triggers it. We don't know what actual part of the brain is causing this hangover at all. We have some lope in the fMRI, and that's about it.

And I think you'd be shocked at how much doctors learn and they know nothing, which actually is sort of like, I don't want to say it's a marginal stupidity, but it is just a reminder of how complex their bodies are, how important it's to listen to them, et cetera.

Chris Gates:
Awesome. Okay, one last piece, and this comes from the space episode. So we interviewed two founders who are working in the space sector, and they had worked together previously at SpaceX. And one thing that stood out to me from that conversation was that at SpaceX, that there was a no acronym rule, and that just acronyms weren't allowed. And that was interesting to me.

Well, one, because acronyms for me, it's like, okay, this is a way that you abbreviate complex terms or complex ideas, and synthesize them into a way that you can abbreviate and talk to people quickly about stuff. But to hear that it actually decrease understanding.

Danny Crichton:
Well, I think we were talking in a recent newsletter about fake analogies in technology. And a fake analogy is an abbreviation, which is a simplified form of a much more complex phrase. And there's research, at least I've seen research, that shows that people have no idea what most abbreviations mean, because the abbreviation becomes the term. And so you forget that there's these languages, okay, LOL, we all kind of know what LOL is. That's a fine abbreviation. There's nothing wrong with it.

The real problem begins when you start going into business terms and you're doing manufacturing and you're like, "Well, let's break the DBD2 system into the RS6," and no one has any idea what that means. Because they start to lose track, because those phrases become more and more diverged and divorced from the original meaning of the phrase that they represent.

So I think what you find is, look, if there were just a couple of acronyms, you're not going to have a problem. But acronyms tend to beget acronyms. And particularly as anyone who's ever interacted with the Pentagon knows, I mean, you get these alphabet soups, we're on a slide, there might be 30 or 40 acronyms, and you're now talking, essentially, a different foreign language. And this is a classic problem in organizational theory, people would rather be silent than to raise their hand and prove that they don't know what something means.

And so a lot of these acronyms, as you get into larger groups, and think about, you're in a manufacturing line, but your marketer comes in, the marketer has no idea what they're doing, but they don't want to make it look like they don't know what they're doing. And so classic organizational theory says, "They're just going to pretend and, 'Aha,' nod their head. They're just going to keep on going."

And as you grow, you end up finding that no one knows what the acronym means. You might actually get to a point a proportion of your corporate population where the majority of people have no idea what the acronyms mean. And so there are some classic case studies, and I want to say, it was IBM in the '80s, that's what's in my head, it may not literally been, or GE or Qualcomm, one of these sort of more venerable tech companies, they created a policy that said you cannot use an acronym, ever.

And therefore, what that did was it did two things. One is you didn't have that sort of [inaudible 00:08:43] of complexity of these terms. And second, you had to get better at naming things. You couldn't just name things these big long phrases, because you had to write them out. And so it actually forced people to say, "Well, what should we actually call this thing?" Instead of the super dupe or hyperactive computer system, it's Watson. And now it's a brand name. And you could argue that that's the sort of abbreviation, but now we're actually talking about it. It's actually an object. It's a brand. It's something that's tenable for a human to understand, and that makes it, I think, a lot easier to communicate.

Chris Gates:
Moving on to our next one, which is, can we talk about this Google researcher who believes the AI he's working on is conscious? It was all over my Twitter feed.

Danny Crichton:
So-

Chris Gates:
Go.

Danny Crichton:
... it went supernova viral and it's just one of those... So there was a software engineer at Google who, basically, started interacting with a model there called LaMDA AI. And as this person was sort of conversing with the AI, prompting it, getting a response back, came to believe that it was sentient. And both triggered an internal conversation around the sentience of this AI, and then was fired because they started leaking screenshots of these conversations supposedly to reporters and outside of the walls of Google, and that's sensitive to Google. And so that triggered a whole conversation of, "Okay, was this actually true?" I think the universal answer from artificial intelligence experts and philosophers was, "No."

Chris Gates:
Was no, yeah, definitely not.

Danny Crichton:
Was no. No one thought it was sentient. But it gets at this challenge around analogies of like well sentience is, as I described it in the newsletter this last week, was duck typing. If it quacks like a duck and walks like a duck, it is a duck. Well, if it talks a human, it's texting back with you as a human, then it's a human. Ironically, the humans don't respond right away to your prompts, that's all part of the game. That's automatically a sign of not sentience, because no one responds that quickly. Some people call it just bullshit. I just call it weak analogy making it.

It's actually trying to simplify complex topics in a way that doesn't actually give you good affordances on what is actually happening in that complexity. The best analogy should be able to summarize something, but still give you a sense of the complexity behind the analogy. And in most cases, we just obfuscate the details to a point that I think a lot of folks who invest in these companies didn't know what they were investing in.

Chris Gates:
So this time you are now arguing that we are oversimplifying. Last week it was overcomplicating. This week is oversimplifying.

Danny Crichton:
Well, they're two sides of the same coin, which is, in order to effectively predict the future, in order to take effective risks, you have to precisely get the right model in your head, right?

Chris Gates:
Right.

Danny Crichton:
You can't be over complicated, you can't be under complicated. You have to be, the Goldilocks principle, you have to have just the right amount of detail where it doesn't overwhelm you, but it has enough detail that you don't make mistakes. Theranos, just to go to a canonical example, was an example where investors invested in this concept, this analogy of prick your finger and I'm going to be able to test everything. Which belies medicine, statistics, the fact that your blood is not homogenous, so that the blood and your capillaries and your fingers is not the same as the blood that's coming out of your heart. The kinds of chemicals and biomarkers that are going to be different in different parts of your body.

And so we kind of deleted out all the complexity of the human body in that analogy. And so it didn't work. And then there's the fraud piece, but the analogy failed. Now, if I were to give you a multi-hundred page study about how complex blood is, that is also not helpful. That is overcomplicating. We would never know invest in anything that-

Chris Gates:
Because no one will ever read, yeah.

Danny Crichton:
No one will ever read it. No one will ever make an investment. So there is a Goldilocks principle of you want to be just balanced between two complex and not too simple.

Chris Gates:
All right, so let's move on to an extremely complex topic, which is China.

Danny Crichton:
No, but there is major story going on. So Congress has been working on this monster bill over the last two years, basically, since Biden came into office in January 2021. The focus has been on, I would call it just industrial policy for the United States, which is how does the United States compete in a variety of frontier technologies? There's semiconductors, quantum computing, biotech, FinTech, the list goes on. And there have been a lot of proposals.

So the House has passed its version. The Senate has passed a different version. In many ways, they don't even overlap. They're in totally different worlds. So I believe the House version is trying to reform the National Science Foundation. The Senate has been trying to increase subsidies to the chips industry. And so there's now a master bill that's going through with representatives and senators all kind of coming together from both sides around a kind of comprehensive bill that would sort of solve the basic science research institutions, and stuff that Sam Arbesman talks about, our scientist in resident talks a lot about, that would also provide subsidies for semiconductor fabrication in the United States. And then a whole litany of other things.

Part of that discussion is about US investors investing in foreign nations, and particularly in foreign adversaries like China. So in the United States today under what's known as CFIUS, the Committee for Foreign Investment in the United States, the United States sort of blocks the acquisition of American companies by some foreign countries. So when a Chinese company wants to buy an American company, that has to go through review. And those reviews were strengthened in 2018 with a new law that was passed by Congress and signed into law. And so it's gotten much harder for Chinese companies and basically any foreign company to buy an American business.

If you want to buy equity, it's harder. If you want to just buy it out of whole, it's actually particularly hard. And that is extremely true within what is known as critical technology. So if a British company wants to buy a semiconductor business, an American semiconductor business, that would have to go through review. It has to go through the Treasury, the Justice Department, and several other agencies in order to get approval. Obviously, it's probably easier if you're British than if you're Chinese because of sensitivities, but that's not to say it's not reviewed.

What's new with Congress's debate going on right now on Capitol Hill is we would actually prevent American venture capitals from investing overseas. So for instance, Sand Hill Road would no longer, or maybe not would no longer, would not be banned, but would have to go through review to invest in any, particularly, Chinese company.

Chris Gates:
So what's your concern with, if this bill passes?

Danny Crichton:
The most important thing is that you want to build up the dynamism and strength of the US startup ecosystem. That to me is goal number one. That is the most important thing. That's what we all want to align around. And so the most important criteria for strong startup ecosystem, in my view, is exits. The ability to make money while building a startup. It's great that some folks will do cashflow businesses or people who will bootstrap. There are other ways besides the high scale, high growth, venture capital model of growing startups. And we have to acknowledge that. But that's not how you build a massive startup innovation ecosystem.

And so when you get exits, big numbers, billions of dollars of dollars in an IPO or an M&A, that attracts founders, that attracts talent, that attracts LP capital, which attracts good venture capitalists out of other industries, where they could probably make money in hedge funds or whatever the case may be. And instead, they're investing in American innovation. And so as other alternatives to acquisitions, M&A get blocked. As VCs are no longer able to invest in certain markets, we diminish the exit value of a lot of these companies.

Take a semiconductor business in the United States, let's say, Intel and AMD, US companies are poking around and saying, "We want to buy it." And a Chinese company shows up and they're like, "We really, really want to buy it. We're going to buy it at a higher price." You could still sell to Intel, but because the Chinese buyer is in the market for that particular company, it drives the price higher, and therefore, the exit value is worth more, even though it's not being sold to that company.

It's just increasing the number of bidders and the more bidders there are, by definition of markets, the price should go up, because there's more demand for the company. I think what you're seeing over the last couple years is there's been law after law that's restricted the number of buyers, that has restricted the amount of venture capital going into these industries. And that just means that it's getting harder and harder in the toughest industries to build startups to grow.

Chris Gates:
And isn't China doing something similar? Aren't they contracting as well?

Danny Crichton:
Yeah. I mean, and in some ways there's a weird harmony or parallel process in which both countries are sort of cutting back on the start of ecosystem. In the Chinese case, there's a deliberate program called Common Prosperity to attempt to reduce some of the inequalities in the current Chinese economy. So to take some money from the big tech winners, namely Alibaba, Tencent, Baidu and others, and sort of redistribute that wealth across the country as well as reduce its power vis-a-vis the Communist Party of China.

Chris Gates:
Yeah, well I remember when they canceled the Ant IPO and that was-

Danny Crichton:
That's right.

Chris Gates:
... such a huge deal.

Danny Crichton:
That's right. And as I'm recording this, it looks like the Financial Supervisory Board in China has just accepted their request to become a holding company, which is the-

Chris Gates:
Oh, no way.

Danny Crichton:
... next block on their IPO. So it's actually, as we're recording this, probably a few days before we publish, it seems like it's a little bit back on track a year or two later. So it's coming. But yes, Ant, which would've been one of the largest IPOs, I think if not... It would've been, I think, the second-largest IPO after Saudi Aramco. It would be the largest in tech, I believe. But Saudi Aramco's IPO has a couple of trillion dollars, because it's a large oil company, there's a lot of oil in Saudi Arabia, would've been the old time high.

So one of the strongest IPOs, pulled back was a huge shock to a lot of folk. Obviously, Jack Ma got kind of pulled out of the public eye for a long period of time, and sort of quietly disappeared or made himself unavailable and hard to see. But that just gets to the point of, look, both countries which have extremely dynamic competitive startup ecosystems are just putting more and more constraints and rules and bureaucracy in place. And to me, obviously, we need to be competitive. We want to beat China. We want to have stronger companies. Restricting access to capital is not a good way to do that.

Chris Gates:
My perception of what's happening in China, which you kind of contradicted in your piece, my view was China's constriction and redistribution was actually stifling innovation. So if somebody were to ask me, "What are your thoughts on startups and innovation in China?" I would say that the Chinese government has made it a really inhospitable place to build a company. Shutting down all of their ed tech. But in your article, I've felt like you were arguing kind of the opposite, that they are actually creating a more rich environment to start a company than potentially the US.

Danny Crichton:
No, I think you're absolutely right. Both countries have put in place restrictions that harm the long-term value of startups. Now, China's actually done far more in the last two years with Common Prosperity, with knocking out the ed tech industry with massively peeling back social media, and a lot of other industries. Where ed tech alone in China was a multi tens of billion dollar industry, and it's gone.

Chris Gates:
It was huge. Yeah.

Danny Crichton:
It's disappeared.

Chris Gates:
Gone.

Danny Crichton:
It's gone. And so in a policy environment where you're like, "I'm going to spend five to 10 years, put a billion dollars of capital to work to go build this business," and then one day it disappears, who wants to go and build a business when it can just be taken away from you, when it can just be shut offline? It's one thing if you're in the United States like cannabis, where it's still a legal market, and you sort of know the risk going in that you're taking this bet that it's going to get legalized at the federal level in the next couple of years. It makes sense.

The US is obviously not putting those kinds of clamps in the same way down. But, look, you're seeing proposals like the upgrade of CFIUS, on investment coming into the United States. This new proposal on Capitol Hill that'll prevent VCs from going out, and not just venture capital, it'll also prevent partnerships, sales, anything that might involve any kind of technology transfer externally to the United States could potentially be under the purview. So you might harm revenue growth, you might prevent joint ventures that are actually quite valuable.

And then on top of that, we have had proposals to tax unrealized capital gains, which is a popular proposal swirling around Capitol Hill, but would massively have an effect on VC and other sort of long term asset classes. And so to me, there's just this inhospitable environment that's getting crafted by DC. And look, China's a huge threat. It's grown massively. It is competitive. Secretary of State, Anthony Blinken, just gave a major policy address on China two weeks or so ago at George Washington University. China's the only country in the world that is uniquely situated to threaten US indemnity anywhere in the world, across all domains: economic, social, cultural, political, international relations, et cetera.

So there has to be a response to that, but you don't want to make it a bogeyman response, where you're like, "We need to change everything we do right, in order to sort of defeat this." Because I think what has always made America strong is its open markets, its competitive free markets, and allowing entrepreneurs to build the best businesses they can. And so long as we're open to that and people are able to come here, they're going to continue to centralize, because of the dense talent networks and the venture capital and the flexibility and open regulations that, again, you build a business in China has taken away from you.

It doesn't happen in the United States. But we're seeming to walk that path, and it's a path that's designed to counter what we're seeing overseas. And in some ways, you have to have the fortitude and the strength to say, "The United States has a good model. It works. Maybe there's some stuff on the fringes you can cut off. But look at the success we've had for decades in our innovation ecosystem. We can't waste that."