Here are more examples of civilzational inadequacy overreaching. Most of them on their own aren’t slam-dunk cases against civilizational inadequacy/mass irrationality, but I hope they at least make a trend visible.
Electric cars in tunnels
I’ll start with a few examples from Eliezer’s post on civilizational inadequacy:
So why doesn’t Earth have autopiloted cars? In the US one obvious answer is screwed-up liability laws, but not all countries are so hesitant. The real answer is that Earth only recently started to reach the AI level required to deal with other humans on the road, pedestrians randomly walking into the middle of the street, deer bounding across the avenues, whatever. But, and this I have not yet heard suggested here, you could solve that problem by having tunnels underground, instead of streets above, and all the cars auto-piloted. Then the AI problem would become vastly easier and could have been solved in the early 2000s of this Earth, if not earlier.
Current cars do not already travel through underground tunnels. Because your cars run on gasoline and would have filled the tunnels with choking fumes, without either (1) a big expensive ventilation system, or (2) expensive electrified rails that would… impose friction costs? Make it too expensive to keep the tunnels in repair? I know that dath ilan just used battery-powered cars in underground tunnels and didn’t bother with electrified rails.
So there is a circular dependency between electric cars that require autopilots, autopiloted cars that require closed predictable tunnels, and tunnels that require electric cars.
This Earth cannot resolve circular dependencies and almost always gets stuck in Nash equilibria.
Another problem might be that tunnels tend to cost $100 million or more per mile. (From here linked above.) Given the massive level of initial investment required, the unclear returns, the long timeline, and the huge network effect that you’d need to make it even remotely worth it, I’d put this down to totally reasonable doubt that the benefits exceed the costs of doing this—especially since the tunnel network looks like it’s going to accelerate self-driving cars by maybe a decade or two at best.
From the same piece:
Around 50 years ago in dath ilan before I came here, our shadarak-trained economists judged plausible the proposition put forth by Sinyelt that the economic friction involved in moving house—e.g. packing and unpacking all your things—was responsible for the economy being slow-to-adapt to shocks. That the friction costs of moving house were a primary reason why people wouldn’t move fast enough to places where the economy was booming, or leave old jobs that weren’t good for them. My home civilization, as you might guess, makes a huge deal out of the virtue of being fast to adapt: fast to respond to facts, fast to update and change policies. And if Sinyelt was right, the cost of people moving geographically was interfering with that virtue. So they built a test city—dath ilan had a concept of “let’s build a test city”—where houses were mated to modular foundations.
You then ask how you move houses from foundation to foundation, especially if, as in nowadays, instead of roads you have underground tunnels built for cars rather than houses. This will take a moment to describe, but keep in mind, there is no technology that this Earth couldn’t handle easily. Imagine the city is divided into a grid of squares 3 and a quarter kilometers on a side. All the grid points have tall narrow posts rising up from the ground, thoroughly anchored in the rock below. All the north-south lines of the grid have strong cables, like the cables this Earth uses to hold up suspension bridges, stretched between each grid point. (The cables are painted blue-reflective so that they mostly fade out against the sky, likewise the tower-posts.) Between two north-south cables, you can stretch east-west cables, side-by-side, attached by movable motors to points along the north-south cables, so the east-west cables can slide along the north-south cables. From points on the east-west cables, you can drop down nine vertical cables, again attached by motors so the vertical cables can slide along the east-west cables. The nine vertical cables pick up a house from its modular foundations, draw it upward, and then the vertical cables and east-west cables slide to convey the house to a grid point. This takes you to an avenue large enough to move houses, which can convey that house to the next grid point. The cables are designed to move gently, and furniture is designed to be stable. Storage things inside the house are designed to allow flexible balloons to inflate inside them to keep their contents secure, and so on.
(Ibid.) Reasons other than “civilizational inadequacy” that this isn’t practical:
Every block needs four “tall narrow posts” that can support essentially the entire weight of a house (about 100-300 tons). Not to mention the incredibly strong cabling and motors you’d need. I’m no mechanical engineer, but this sounds like the engineering nightmare to end all nightmares.
This plan asks for airbags in every storage device and “stable” furniture that doesn’t move or fall over when it’s winched up and moved over. Now the test city needs specially-made versions of everything you might put in a house, which means residents have way less choice over what you buy and way less economy of scale in producing it.
Every time part of this system messes up, you drop a house on another house.
The average (wooden) house weighs 30-60 tons, I’m pretty sure nothing except possibly some sort of nanotube cable could hold that plus its own weight over 3km. Steel certainly can’t. Not to mention the towers to support the cables, which must be simply colossal, each one will be under a minimum bending moment of ~4.5x108 N, which wolframalpha tells me is the force water exerts on the walls of hoover dam, which gives an idea of scale.
Opt-out organ donation
How about this one from AGI Outcomes and Civilizational Competence.
getting good outcomes from AGI looks much harder than the kinds of things we routinely fail at, like bothering to switch to opt-out programs for organ donation.
Actually, opt-out organ donation isn’t a very big win, if it’s a win at all:
In terms of opt-out, I only know the data in the U.S. but basically while it might be a good idea, it’s unlikely to yield significant increases: it seems like such an attractive decision architecture/nudge type intervention, but when you dig in, it’s a much closer call (which is why Sunstein and Thaler don’t recommend it, for example).
There’d also be some risks of going to an opt-out system. So imagine what happens when someone is actually eligible to become a deceased donor — they’re brain dead, so their death is often sudden and unexpected to their family. Let’s say the donor didn’t choose to opt out. What do you tell the family who’s in the room with a brain-dead patient whose heart is still beating — “she’s dead. Now you have to leave the room so we can harvest her organs”? If they have no control over what happens to their loved ones body, you can imagine a lot of people becoming pretty upset, even people who might have been persuaded to say yes to donation. The deceased organ donation relies on the public’s support and good graces, so if you have repeated instances of grieving families publicly decrying the opt-out system, that creates a significant risk that the change will be counterproductive.
Thanks to Holden Karnofsky for pointing this one out.
Another person who often mentions civilizational inadequacy is Peter Thiel. For instance, he argues that Solyndra, a high-profile failed publicly-funded solar power company, was doomed from the start because of basic math that government funders failed to notice:
In terms of investing in science and technology, it seems to me that the minimum criterion for doing it is to have some understanding of these things and some ability to evaluate them properly … I always use the Solyndra bankruptcy as an example in this question of what went wrong … A mathematical objection to it was that a cylinder has 2πr the surface area of a flat panel, which would be 2r and therefore is, by definition, 1 over π as efficient as a flat panel. You could just use ninth-grade, high school geometry to show that this was a demonstrably inferior technology.
It was never going to be commercially viable. You have a Nobel laureate, Steve Chu, running the Energy Department who is not allowed to use ninth-grade high school geometry in evaluating what to do. That sort of a society, that sort of a government is one that should not be allowed to make any investments in these areas whatsoever.
(source) In other words, nobody noticed that Solyndra’s product wouldn’t work out, as a matter of basic math, before investing $1 billion in the company.
Now, this does occasionally happen—see, e.g., the dot com bubble. But often there’s another explanation. In Solyndra’s case, for starters, Thiel’s claim that the cylinder design was 1/pi as efficient as a flat pane only holds when the sun is shining directly down on the flat plane. If the sun is angled at 90 degrees, for instance, then the cylindrical cell gets just as much sun (if it’s not in the shade) while the flat panel gets none.
To mitigate this, flat panels often come with motors that move them around so that they capture more of the sun. But this equipment doesn’t work very well in high winds, and it takes up space, leading to lower power density per roof. If there are high fixed costs per roof, then sunlight-per-roof matters more than sunlight-per-panel anyway.
So the selling point of Solyndra’s panels was that they didn’t need anchoring equipment and motors, meaning that you could fit more of them on a roof, installation was less expensive, they were more robust to wind, etc. They traded these benefit for additional manufacturing costs, because they required a special surface area coating and the cylindrical shape was harder to manufacture. A priori, it doesn’t seem like this is a trade-off that could be resolved by ninth-grade math.
And indeed, the efficiency concern wasn’t what ultimately sank Solyndra. Instead—at least according to Wiki and some other articles, the problem was their specialized coating. While Solyndra was scaling up, standard solar panels were simultaneously getting way cheaper. But Solyndra couldn’t take advantage of this because they needed a specialized manufacturing process to make them cylindrical. So they lost on price, but not at all for the reasons Thiel was suggesting.
Solyndra probably had a bunch of other problems as well. My point isn’t necessarily that it was a good investment, or even that you couldn’t have figured out a priori that it was doomed. They probably scaled up too quickly, and maybe the trend towards cheaper traditional panels could have been figured out in advance. And plus there were a bunch of accounting/bookkeeping scandals, although those probably were effects, rather than causes, of the company floundering.
But I think it is clear that the flaws in their plan couldn’t have predicted by 9th-grade geometry, and that it’s not too hard to call that into question if you do some research into the actual technical claims they had. So if you find yourself arguing that government incompetence has reached the level where “a Nobel laureate… is not allowed to use ninth-grade high school geometry in evaluating what to do,” this should probably throw up a red flag that it’s time to start doing some of that research.