Links #11

Solar EROI, printed houses, synthetic cattle feed is surprisingly efficient, and more

Besides the vignettes, I’ve bolded some of the links I find most interesting. Much of the other stuff I put here is for future reference and might not be as interesting to read.

1.

The paper A Polysilicon Learning Curve and the Material Requirements for Broad Electrification with Photovoltaics by 2050 has a nice chart of the energy use for solar cell production:

Since solar cells produce energy, we can think about this in terms of an “energy return on investment” or, in other words, the number of wafers you can produce with a single wafer. I estimate that it’s 20x even with conservative assumptions. The new Longi cells discussed in a previous linkpost reach 180x!

This is a fundamental limit on the growth rate of solar cells, but it’s not much of a limit, a single wafer growing at the associated annual rate would cover the entire earth in 17 years.

To get another handle on these numbers, let’s think about the energy density of solar silicon. By my count it’s 28 GJ/kg, about 1000 times higher than gasoline. Silicon is the new oil!

If I use the new Longhi numbers, I get 244 GJ/kg, a mere 312x away from the incredible energy density of fission fuels like uranium. It’s not really a fair comparison since the energy in uranium is exhaustible while solar panels can presumably be recycled, but interesting nonetheless.

Source

Related: a nice site on how photovoltaics are made.

2.

What Makes Housing So Expensive? Great piece with lots of data on construction prices and land values. Notice that costs are finely divided between a lot of different factors, a sign that the process may already be pretty optimized. I hope someone will finally figure out prefab construction, but many companies have tried and failed.

Brian Potter has other pieces on companies trying to make housing much cheaper:

Toyota's Prefab Homes

The Katerra Team Rides Again: ONX Homes

What Progress Has Icon Made on 3D-Printed Homes?

Icon’s 3D printed homes idea is fun because you could potentially roll a truck up to the job site and have it print a house on it’s own, the only operating cost being the price of raw materials. For example, BBC ran a piece on a 3D printer big enough to build a house.

There’s two challenges here. First, getting a 3D printer to work in the real world from the back of a truck is extremely challenging. But more importantly, homes are far more than just structural material; they need insulation, plumbing, wiring, and windows. If the frame is 3D printed but wiring still has to be done by hand then construction will still be bottlenecked by the speed and cost of labor.

People are working on 3D printers that can weave together multiple materials like wiring, but until every stage of construction is automated it will be hard to compete.

Related: the company Orbital composites is 3D printing all sorts of stuff for space. Can we get that kind of sophistication on the scale of a home for a reasonable price?

3.

We’re getting closer to making food from air. The company Calysta makes cattle feed from natural gas and the company Solar Foods makes single-cell protein by feeding microbes hydrogen. These ideas aren’t that new: Methylococcus capsulatus has been used for decades to produce cattle feed from natural gas.

But these efforts become more interesting in the context of companies like Terraform Industries that make natural gas from air. The methane can be used to make cattle feed, skipping plants entirely¹!

The weird thing is that synthetic cattle feed plus dairy cows is already as energy efficient as plants themselves. Solar panels are about 25% efficient, I would guess Terraform’s process can hit about 30% efficiency, conversion to feed is maybe 80% efficient, and dairy cows produce milk calories with an efficiency 24 percent. Taken together, that’s an efficiency of 1.4%, similar to typical crops.

Even better, the output is milk, a much more valuable product than raw crops². The whole feed manufacturing process can happen on the feedlot, eliminating shipping costs. The land use would be much lower than growing feed, which is a huge portion of overall land use.

Everything else

Why I No Longer Prioritize Wild Animal Welfare. The author argues that farmed animal welfare is more cost effective in the short term, that it’s hard to convince people or governments to take action on wild animal welfare, and that in the far future other beings like AI’s or brain emulations will be more numerous than wild animals.

Labor, Capital, and Patience in the Optimal Growth of Social Movements. A mathematical model of optimal social movement growth. I think this has implications for pretty much any social movement. It’s probably not worth reading on it’s own but I want to do a more in-depth post explaining the math at some point.

Beyond Bitcoin's Hype, Real Use Cases in Africa. I’m going to continue pointing out that in some ways developing countries have more interesting financial institutions than the west.

How Should We Think About the Strategic Petroleum Reserve? The theory of exhaustible resources says that they should grow in price equal to the real interest rate. When states hoard exhaustible resources, they add a new risk: the state might suddenly decide to sell their resources and crash the price. Because of that risk, the price of exhaustible resources has to rise faster than the real interest rate. There are other policy implications discussed in the post.

Rebuilding US Arms Production - Can a new Strategy Restore the Arsenal of Democracy? Review of US arms production. Argues that US can restore peace by out-producing enemies like Russia, Iran, and China and that this is straightforwardly possible.

Causal Analysis of Policy Effects on Fertility. A large review of fertility interventions, many seem to work and have lasting effects.

Extracorporeal cardiopulmonary resuscitation. Basic innovation in medicine continues with this technique that oxygenates a patients blood during cardiac arrest. Given how many people die of heart attacks this is pretty important.

Vague premonitions about the RNA memory hypothesis. Initial results finding RNA molecules that stick around nerve cells for a long time.

Gordian Bio is touting a technique to randomly place different gene therapies into different cells in a living tissue and then sequence individual cells to measure the effects. Their goal is to target age-related disease. This is a pretty neat idea, though I’m not sure if harvesting the tissue requires killing the animal they are testing on.

Sculpting photoproducts with DNA origami uses the folding of DNA to spatially arrange dye molecules. I’m skeptical of the value of their proposed applications but using DNA to position other molecules could have other cool applications like in lithography.

μeV-Deep Neutron Bound States in Nanocrystals. This is a theoretical paper, but having free neutrons in a material could have applications for fission or fusion energy production.

DiLoCo: Distributed Low-Communication Training of Language Models. It’s this kind of stuff that makes me think solar-powered datacenters could become a reality. If you can distribute AI training runs and keep capital costs low, then training can run only when sunlight is hitting your datacenter. Austin Vernon is skeptical has a nice piece on how we might run datacenters on renewable energy.

Sciama’s argument on life in a random universe and distinguishing apples from oranges. A short piece giving some intuitions for high-dimensional probability and how that relates to the Fermi paradox.

High-Density Days is a concept for building friendships. Since friendship is partially a function of how much time you spend together (and the quality of that time), a few infrequent adventures can cement a friendship better than frequent, short hangouts.

1. Of course, if we’re going to synthesize feed why not just take the extra step and synthesize food directly? This is a long way off but I think it’s doable.

2. And we’re ignoring the fact that only a fraction of a plant’s mass is edible and needs further processing.