A response to splitting infinity
Tech he's skeptical of, and what I think
I wrote a long response to this
and then substack deleted it. So here we go again. This time it’s an article, because it’s too long to post as a comment.
> Back-of-the-envelope calculations are inherently optimistic.
Not always. Sometimes your back of the envelope calculation involves an existing cost. But that tech is currently only used at small scale. If you 10x the worlds electrolysis for hydrogen production, you can expect someone somewhere to figure out a cost saving compared to today’s electrolysis.
> e-Fuels
Might make economic sense under some conditions. Especially with strict carbon credit laws and geopolitical energy security concerns.
> Capturing carbon and making fuel is probably more expensive than just capturing carbon.
You need to compare capturing carbon + making it into fuel. Vs capturing carbon + long term storage of that carbon + digging different carbon out of the ground + refining that black sludge into usable fuel + transporting that black sludge from a petrostate dictatorship to where you want it.
> Launch costs below $100/kg to LEO
Agreed
>Asteroid mining
Agreed
> Space data centers
Agreed
> Ramjets and demand for Mach 3+ flight
This depends very much on how much rich people are prepared to pay for a quick journey.
> Passenger rail
There is nothing wrong with steel wheels on steel tracks as a technology. The current issues with passenger rail are that the rail line doesn’t go to your door, and that changing trains is inconvenient. Theoretically we could have steel tracks to your door, and 1 person automated “trains” that take you to your destination. Practically regulations and switchover costs probably stop this.
> If you want to go long distances quickly, an airplane can get you there faster and cheaper.
Airplanes aren’t energy efficient, and they can basically only use petrochemical fuels for long distance. If you have super cheap e-fuel, aircraft make a lot of sense. But if your fuel is only getting more expensive, the train looks like a better option.
> Fusion
Could potentially succeed by being less regulated out of existence than fission. May also have the potential to transform mercury into gold with high energy neutrons.
> Non-solar energy sources
Theoretically thorium could be very cheap. In practice, who knows. Regulations and or some other issue may well stop that.
Quantum computing
> We already have good methods of simulating chemistry on classical computers.
I wouldn’t describe current methods as good. Somewhat adequate, maybe.
> Gene therapies in humans
In the short term, this will focus on fixing monogenic disorders. In the longer term, all sorts of wild transformations to biology are theoretically possible, if you can redesign genomes wholesale. At the very least, turning people into centaurs or mermaids.
> Worse, consider all the modalities that might substitute for gene therapies: RNAi, RNA-targeted drugs, vaccines, antibodies, CAR-T, in-vivo CAR-T, stem cell therapies, and 3D printed organs. Each has an advantage over gene therapies in some aspect, and many are likely to be cheaper.
And I could say the same about these technologies. There are many different approaches, each with their advantages and disadvantages. This may be a reason to be a bit less all in on one approach, but it makes the whole system more powerful. And their might be synergy, like a piece of RNAi that prevents errors when changing DNA.
> Brain-Computer Interfaces for healthy adults
If it’s cheap and safe, some people will want it. For full immersion VR porn if nothing else.
> But it’s not clear that healthy humans can use BCI’s to output far more information than they already do.
When I try to visually imagine something, I can imagine a picture in far less time than it would take to draw that picture in image editing software or describe that picture to an AI.
For well thought out text, typing is probably near optimal, if a buggy substack UI doesn’t delete your post and make you retype it.
>”Nanotech”
> These fields will continue to change the world. My skepticism lies with the nanotechnology that remains.
Imagine it’s the 1950’s. There are computers that consist of 100’s of vacuum tubes. And scifi descriptions of computers as magical machines that can answer any possible question. You might say “but we refuse to call adding machines AI because it already exists. My skepticism lies with the AI that remains”
Imagine a future industrial process. How do you distinguish between “really advanced chemistry” and the nanotechnology you are skeptical of?
> Early visions of nanotechnology had tiny robots that could carry out precise tasks. In harsh conditions and high temperatures, such robots are infeasible;
I am not convinced of this. But if nothing else, such robots aren’t easy.
> If you’re willing to accept standard conditions, proteins are a remarkable nanotechnology that can perform almost any desired physical transformation. No need to invent a new field when biotech can do it all.
Proteins are something that evolution came up with. They are probably not the optimal way of arranging atoms for manufacturing, they are just some solution that works well enough.
There are plenty of things we can’t currently do with proteins. For example, produce diamond fibers. There are a whole lot of impressive new capabilities that are theoretically available and not currently used. You might use proteins to do this. But proteins are just one molecular structure out of many, and probably not the best choice.
> For stuff we have today, it makes more sense to build it directly than to try to build it with nanotechnology
The stuff we have today has been designed to be easy to build directly.
> Say you want a car. You could feed the bots some metal, and they could assemble a car, but isn’t that just a factory?
Well if you have a desktop device that can produce almost any currently manufactured product (that fits in the box) out of raw materials, fully automated, that’s very different from any existing factory. It’s closer to a 3d printer, but for everything not just plastic.
> So maybe the nanobots go out into the environment and gather metal. They need to be self-repairing and have some sort of search function and we need some way of beaming energy to them to break chemical bonds.
> Instead of tackling dozens of near-impossible technical problems, you can just build a car the old-fashioned way: processing many atoms of ore simultaneously, arranging many atoms of metal all at once.
Potatoes are cheap. Potatoes are grown using something that looks a lot like self replicating nanotech gathering materials from the environment. The technical problems aren’t easy, but they are one off costs. Once you have the nanobots, you can keep building more stuff at hardly any marginal cost.
> Perhaps there are some things we simply can’t make with current techniques? Not really,
All right. Make me a 1 meter cube of diamond, such that, when you number the atoms, the prime numbered atoms are carbon 13, and the composite numbered atoms are carbon 12. Atomically flawless, no mistakes. Actually, leave a small gap in the middle. just enough space to fit a molecule of a ridiculously unstable nitrogen compound. Now all your manufacturing needs to keep that molecule below -20 or it degrades.
> For computers, one of the few applications that requires nanoscale precision, lithography will take us to the limit of what’s possible.
Computers, one of the few applications so high value that lithography makes economic sense. Lithography struggles with single atom precision. It’s slow and expensive. It’s limited to applying thin layers, so you can’t make 3d structures. And if you use x-rays for the precision, you fry any sensitive molecules.
> It’s in lithography that the last holdout of nanotechnology, atomically precise manufacturing, shows the most promise.
If you have anything resembling good atomically precise nanotech, you sure aren’t using lithography. You are making computer chips far faster and cheaper than lithography ever could.
> The throughput and equipment costs are too high to replace industrial chemistry,
One thing that atomically precise manufacturing should be good at making is more atomically precise manufacturing equipment. I suspect that high value chemicals are likely to be the target first.
> Structural materials
There are, like diamond and boron nitride and sapphire, various materials that are made out of common elements, but are currently too expensive and hard to make to be practical.
> And then there’s wood, a material synthesized from air and sunlight
Do you think that trees are magic? There are a huge number of possible carbon compounds. Trees are just something evolution happened to produce. Cellulose is just one organic molecule out of many. We don’t yet have the ability to design something like that ourselves, but we will.
> For bulk structural materials, I don’t think we’re going to beat these three on cost per unit of performance any time soon. They’re local maxima that take advantage of available resources.
So no genetically engineering trees to produce silk or diamond or something? No cheaper way to refine titanium?
> For alloys, we can only paint with the periodic table, and we have to make these out of Earth-abundant elements while obeying constraints on what elements alloy with each other. Much of this space has been explored,
Well yes. We have mostly explored the space of taking various elements, sticking them in a pot, and thermal cycling them. But there are a huge number of molecular structures that these elements could theoretically be arranged in, and most of these structures can’t be achieved by a process as crude as heating it up for a bit then cooling it rapidly.
> High strength-to-weight materials
> You pretty much have to make it out of a carbon allotrope or some sort of glass.
There are a LOT of carbon allotropes. And we lack the ability to manufacture them well. Cheap high quality diamond fibers or nanotubes would change a lot.
There are also boron fibers, boron nitrides, silicon carbides, etc. And there are various highly speculative options like magmater.
Oh and silk has a pretty decent strength to weight ratio. It could be good if you could figure out bulk protein synthesis.
> Electrical materials
High purity carbon nanotubes beat copper here.
> Catalysts
I thought you said that proteins could do everything? Yet they aren’t mentioned here?
> The same issues are repeated across chemistry and materials science, namely that we’ve found cheap and effective solutions in most applications.
When a solution is first invented, it usually doesn’t start cheap and effective, it starts expensive and/or limited. Then it gets better. A lot of this feels like “I can’t imagine something we haven’t found yet”.
> The frontier of materials science looks more like “try lots of existing materials and processing conditions for your specific application” rather than discovering some new composition on a computer.
Which leads back to the importance of quantum computing, current algorithms are not that great.
> Economics of home robots
> but if robots are very competent they make human labor cheap.
If you think of the spinning Jenny as a simple robot that only does one thing, then so far we have seen more automation leading to a higher value on human time. This may change.
> So you need a weird situation where human labor is super scarce but also replaceable with robots but only in the home.
We might have a situation where everything can be automated. Lets optimistically assume AI alignment is solved. Then the AI gives us home robots, because we don’t want to tidy up and the AI is nice.
Home robots make sense whenever there is some economically valuable task that only humans can do, but picking up dirty socks isn’t it. And picking up dirty socks doesn’t seem to be the pinnacle of human performance.