Discover more from Strange Loop Canon
The next S curve
Ways that we might say goodbye to the Great Polarisation after all ...
[Part of an ongoing series - one, two, three, four, five, six]
I'd written before about how the Great Stagnation thesis has been an observation about a stasis, of reduced returns to increased efforts, across a wide array of sectors from science to technology to organisational productivity and manufacturing. The interesting corollary that I've been exploring is how this creates a Great Polarisation within the world too.
The thesis I've been developing has been that there's been a drop in productivity across the board in most areas of life and work. This slowdown has been the result of our phenomenal ability to get in our own way through regulations, risk aversion, bureaucracy and the general awfulness of coordination tax.
But the fact that we live in ever more complex hierarchies means there are tradeoffs - that the costs we have to bear also increases. It's what creates some version of resiliency while also decreasing the returns we get from pursuing the same scaled efforts as before.
It's the reason we spend most of our time in maintenance mode as organisations get larger. Not only is that inevitable in some sense, it's also necessary to create a stable base if we want to grow thereafter.
We've tried to hack our way out the reduced effectiveness and rising complexity conundrum through increased specialisation. After all, that's proven to increase our ability to get better at doing things, and is almost a bedrock principle of economics. But this also reduces autonomy at an individual level and increases the Great Polarisation.
But where's the end? The argument that we're in a Great Stagnation implies by its very statement that this is an aberration. Is it though? My thesis is that we're in a period of stasis simply because the number of technologies we need to progress don't all progress at the same rate.
The previous wave somehow coincided, plus or minus a decade, on our ability to get way more juice out of ongoing research in a phenomenal array of things - transport, power generation and distribution, medicine, rockets, <insert your favourite innovation>. This coincided with (or is causally related to) our ability to use new business models of automation, outsourcing, offshoring and digitisation to truly kick ass and make the growth rates itself grow. Everyone prospered.
But the fact that we've not been able to make fundamental breakthroughs we like in physics, or discover nanotechnology, or solve cancer, is not a knock necessarily on our abilities. It's also because we've been stuck in a paradigm slightly longer than we were in the past.
And progress needs time to catch up. Sometimes we need improvements in chemistry to spark physics, or biology to spark chemistry, or Greek mythology to spark biology. Sometimes we need one thing to grow a lot before we're able to make use of the new tools that emerge to grow other fields.
And it looks like we're reaching that point with the growth of automation and computation. Growing out of the Great Stagnation is possible with this sequence of affairs, as it was always wont to do.
So this is a natural pause point of the Great Polarisation series. With a look at the ways in which the stagnation could end in the coming decade. These are not necessarily examples of what would restart the productivity by removing the obstacles that we've constructed in the way, through design and as fruits of our success thus far, but probably by finding ways around it.
It's not a panacea and it's not forever, but it'll probably be enough for the next S curve. And most importantly these are all enabled by certain platforms which themselves have been in development for a decade or more. And it's through the judicious application of several combined technologies that we'll start to see the curve bend up again.
Will this be enough to reduce or eliminate the Great Polarisation? That remains to be seen. For that we need to not just look at the increases in averages due to productivity but also its distribution. The distribution is bent by regulations and governments, but also by the collective need to utilise everyone's abilities so that their labour becomes useful again.
Onwards to speculation across several dimensions.
Since the days we stopped chilling as hunter gatherers we've been working on agriculture. And it's an enormous pain in the ass. Since the industrial revolution we've had movements that reduced the number of farmers even as we've vastly increased the total output. This is about to change again.
The concept isn't strictly about farming only. But a chance to retool the entire system. Today if I want lettuce somebody has to plant it, grow it, ship it and sell it to someone who'll sell it to me. Even more than the waste in time and money, it's a quality bust at times. As someone who's been hunting for good mangoes in London for a decade, some things aren't meant to be.
There's farm to table, which reduces middlemen. Still doesn't manage to remove geography from the equation. And that's a problem.
Not just because sieges are lost on food supply or because armies march on their stomach. But also because if food is grown closer to consumption that saves time, energy, wastage, money and quality degradation.
Why couldn't we do this before? Because vertical farming is annoyingly specialised and tough to pull off. The construction industry has barely managed to make itself more efficient, now can you imagine telling them to also grow stuff inside their concrete jungles?
The ability to do this comes from being better at robotics, irrigation techniques, replicable knowledge about crop cultivation (unlike those pesky Balinese farmers and their religious methods) and AI to help manage everything. That's rise in computation and automation right there.
Which is how we have the situation where 2 acres can out produce 750 and with 95% water savings. If nothing else this should stave off the water wars.
Biology as a platform
Which displays Banking as a service to become the true BaaS. Medicine has always been single shot. It's annoyingly imprecise and difficult to replicate. We've seen a few innovations before in identifying varied existing biological species that helps bring, but we're finally getting at modifying the source code level. This is possible now not just because of changes in the potential analyses, but also because we have far better techniques to test and simulate biology.
For mRNA based vaccines for example, we had to figure out how to make sure they don't screw with the immune system, encourage some occasional moves by the immune system to fish the mRNA from the blood. We also figured out how to get the mRNA some protective covering, by putting them inside small capsules so it doesn't get harmed. From a 2018 Nature paper:
mRNA vaccines represent a promising alternative to conventional vaccine approaches because of their high potency, capacity for rapid development and potential for low-cost manufacture and safe administration. However, their application has until recently been restricted by the instability and inefficient in vivo delivery of mRNA. Recent technological advances have now largely overcome these issues, and multiple mRNA vaccine platforms against infectious diseases and several types of cancer have demonstrated encouraging results in both animal models and humans.
But to do all this you also needed to get good enough to quickly and accurately determine the genetic sequence of a virus, itself something that has been 1000sX price drop in the past two decades.
The possibilities here are that we will finally be able to play around with source code within our cells for all the good and ill that brings. But in either case the fact that we're getting to better understand the biological world we inhabit is a massive plus!
Starting with asteroid mining. Not particularly useful, most likely, for terrestrial benefits. But the path to get there might be beneficial in and of itself. The NASA program to put man on the moon helped bring about a whole array of benefits to us. In medicine, flight control systems, earthquake proofing, hearing aids and more.
While this is highly speculative, the fact that its being seriously considered by both governmental and private companies is indicative of the ambition here.
There's also the more, dare I say, down to earth ambitions of satellite launches and manned space flights, which are all great indicators that there's a new frontier that's shifted from pure scientific exploration to better commercial exploration. And with it comes efficiencies.
Cheaper and better energy
Through both cheaper and more efficient renewable energy sources and better battery technology there's potential to finally bend the Henry Adams curve after five decades of stasis.
There's also the matter of ongoing research into fusion and more portable nuclear reactors.
An exponential technology, these batteries have been dropping in price for three decades, plummeting 90 percent between 1990 and 2010, and 80 percent since. Concurrently, they’ve seen an eleven-fold increase in capacity.
From the World Nuclear Association:
About 100 power reactors with a total gross capacity of about 110,000 MWe are on order or planned, and over 300 more are proposed. Most reactors currently planned are in Asia, with fast-growing economies and rapidly-rising electricity demand.
And showing just the ones with 2021 start date:
Equally interesting is that the UK is spending $250m to help bring about the world's first fusion power plant. Again, the success here is not the point as much as the effort. This follows the footsteps of China and a European one.
And there are also several companies around the world that are trying to make nuclear fusion reactions more portable. There's CFS that's MIT backed, TAE in California, First Light Fusion in Oxford, Tokamak Energy in Oxford, and the world's largest fusion project in ITER in Southern France.
This is a technology that's always been a couple decades away and still might be so. But anything aiming to bend the Adams curve is worth noting.
This is not mentioning the stupendous progress we’ve made in the storage of energy either, since that unleashes an entire ecosystem of “capture to use later" energy systems like solar or wind.
The reduction we see in this Nature paper directly points to several technologies all getting cheaper, and therefore more ubiquitous.
A bit of a longshot even by the rather loose standards of utopia-hunting, but something that can change what we think we know about computation possibilities, physical systems simulation and cryptography.
It's unclear how far along we are here. On the one hand it seems clear that there's demonstrable progress in the space. On the other hand extrapolating from what seems like a good test outcome to potential earth shattering scaling up is a tough sell.
Self driving cars
There's way too much speculation that abounds here. But suffice it to say it'll be pretty cool when it happens, even in a limited sense, and that's coming up for sure in the coming decade. Enough said.
One major initiative started during Obama's term was the Materials Genome Initiative, aiming to speed up the pace of innovation in material science using, you guessed it, open source methods and AI.
The project aimed to map the millions of possible combinations of elements (remember your periodic table people, there are quite a few of those) and make a giant database that scientists can play with. This allows us to simulate and test novel ideas with a real map of the world, and then work on new fabrication tools to make it work better.
It's unclear whether it will allow us to go the whole hog towards nanotechnology. My bet is that it won't, because scaling down 100x in size while retaining fidelity is not what you'd call an easy problem, but any step in that direction is a right step.
This also has a dependency towards the Cheaper energy section before, since better materials vastly improve lithium-ion batteries too, which are ubiquitous in the world.
Yet another use here, perovskite, has the potential to double again the efficiency of solar panels.
It's often overlooked because, well, it's not nearly sexy enough. Unlike Mrs Robinson, we don't think the future is plastics. But maybe we should pay a bit more attention because a breakthrough here affects most other industries.
The joy here is not just about the potential to reduce animal suffering, though that's immense and noteworthy on its own, but also to see where our level of ability to create foods we want from scratch takes us. Even a non-Singerian argument is interesting here. It's the first time really that we've been able to make food en masse that's fully substitutable for something with 10x the carbon footprint.
They're not 10x healthy enough compared to old school meats. Yet. Right now they hAve all the vitamins and minerals you'd like, but make up for it by being heavily processed and high in saturated fats. But that's a curve that will inevitably get bent. With lab grown meats and near substitutes abundant, with substitutes also for chicken and other flavours also resurgent, there's just a sea change coming to the industry.
Is this going to be a footnote in innovation trends regarding agriculture and food, or a star player? It's unclear. There's likely to be some debate on this for time to come, and the markets sure seem to love the path forward, as reliable as an indicator as that is.
Deepmind has been on a tear. They started by beating IBM's amazing Deep Blue records for chess, which quickly got written off as "of course, it's a complete knowledge game so makes sense a machine could beat everyone."
Then was Go, then Shogi. Which also got written into a similar narrative of "sure, total information games yea. That's what computers were meant to do."
Then came heads up poker where ___ has been annoyingly successful. Sure, maybe that's just better playing of the odds and some clever programming on how and when to bluff.
Then StarCraft, where AlphaStar won against Grandmasters and ranked amongst the top 0.2% of players. The excuses were slightly different here, it was that the computer, though restricted to only a few moves at human speed, could see the whole board. And there were the usual arguments that were advanced again of how StarCraft is still an easier challenge than real life because while it has incomplete information, the rules are still known and the world is computationally tractable.
And most recently, MuZero. The papers are still being written, but it seems like the learning component has been sped up in this version. And the implementation has even gotten slightly easier. In the helpful and vague illustration that DeepMind has provided everything seems to be moving in the right direction.
The key here is that MuZero has the ability to focus its attention somewhat by creating a model of its environment with extra attention on the details that are more likely to be salient. In a turn of phrase that someone clearly loves, because it crops up literally everywhere, "knowing an umbrella will keep you dry is more useful to know than modelling the pattern of raindrops in the air".
There's Open AI too that's jumping in the natural language generation game with ever increasing sophistication in their algorithms. GPT-3 seems like an even bigger success than GPT-2, as numeric advances should be, and has created genuine progress in the generation of generic text. While it won't write War And Peace for you yet, Copy AI has made copywriters obsolete. And if that isn't an unqualified good, what is?
When looked at in isolation it feels like small advances within an old paradigm desperately clawing back power from using extreme compute power. But it's also important that the combination here is what's interesting. Finally in AI it feels like there's enough heterogenous vectors that researchers are attacking it from that at some point in the not too distant future there's likely to be attempts at convergence. And it's when some of these things start getting brought together that the true productivity (that's been predicted for a few decades at least) is unleashed.
This deserves an article on its own. But ever since we've been fascinated by the emergent properties inherent in turbulence, in thermodynamics, in material science, and with the studies of phase transitions, we've been focused on how parts of a system might differ from the whole.
It's an understood topic, even banal at times, that there are outcomes in a system that aren't predictable due to the interplay of its components. And turns out they're everywhere. In economics, physics, biology, computing and AI, sociology, even traffic management.
We've had limited tools to analyse the behaviour, much less knowledge on how to use it. But that's changing too. While the dream of analytically predicting the outcomes is unlikely to happen, as is an psychohistory, the possibility that we'll start getting better at managing larger complex systems is no longer a complete pipe dream.
Whether it's through the work of Santa Fe institute or the efforts of heterodox researchers across academia, there's both simulated and heuristic laden approaches being developed to help us work through this.
If nothing else it might help us manage organisations better. And if that isn't a General Purpose Technology then what is.
In the end..
The argument here is that various strands of research and technology that's been stuck for a while are starting to come together. Quite often it's because several fields have gotten to maturity together at the same time (material science meets solar panel efficiency to get to cheaper energy, AI advances enables drug discoveries, CRISPR and analytics progress enables biological platforms). Looked at individually they might seem like they're peaking or troughing or plateauing.
But we forget sometimes that that's not the right way to look at it. And TFP isn't a godlike metric, it's a backward looking measure.
Some of the stuff above will make people very rich. Some will help the society prosper! And some will make the economy grow faster. But most importantly some of them will also reduce the polarisation that's been felt by several sections of the workforce. But the fact that several skillsets are simultaneously becoming the flavour-of-the-decade means that the secondary effects that'll reverberate through the economy will be much stronger!