Doom is Unsustainable


I've got a little motor called ATP synthase. It converts glucose to adenosine tri-phosphate with greater than 85% efficiency. Indeed, I have hundreds of those little motors in every mitochondria, and tens to thousands of mitochondria in every cell in my body, busy converting too much glucose into about 1e24 ATP molecules a day, then into the spare tire around my waist.

As I write this, I am looking at a 50 meter tall Douglas fir in my Oregon back yard, one of many. It also uses ATP synthase, sometimes in reverse, and rubisco and chlorophyll and many other molecular machines to turn sunlight into chained sugar cellulose, building a pile of wood and roots and bark and needles that weighs a thousand times what I do. That tree is a gigantic complex of molecular machines, vastly more elaborate than our entire industrial civilization. Such machine complexes comprise the vast bulk of the biosphere - humans are a few pollinators with childish attitudes. Gaia keeps us around because a few of us are learning to expand her home, out to the stars.

The sun churns out 380 trillion terawatts, almost all of it wasted in intergalactic space, never to touch matter again. The Earth intercepts one ten billionth of that, and life captures less than a fifth of that. Opportunities for growth.

Meanwhile, humans have made about 1e21 transistors up to 2015. The number doubles every 18 months. My friends in the device lab at Intel are researching the next 10 generations of Moore's Law, and have computer models of single atom transistors. They don't know how to make atomic transistors, but have 60 years to figure it out, more if the market cools down a bit and gives them time to think.

Intel races to be first, their competitors race to catch up, a Red Queen's race that has expanded the semiconductor industry 5,000x since Apollo went to the moon with the first computers made from integrated circuits.

That has already caused vast ripples across the entire supply chain. Your 80 gram smart phone replaces an 80 kilogram desktop computer, display, keyboard, UPS, etc., and connects you to a planetary web of data center supercomputers, shared with billions of other web-connected people. This continues a trend that has been accelerating since the 1960s, doing more with less. We make a ton of steel with 20% of the energy that we used in 1950, to make lighter/cheaper/more-efficient machines. As my body and the fir trees demonstrate, we can go much further than we have so far.

I use very little electricity, and do not voluntarily connect myself to wall sockets. My machines use kilowatts, but machines can be vastly improved. Intelligence and knowhow substitutes for resources, which is why looking at the flow of bulk materials is very misleading (and volatile commodity prices measured in variable dollars, more so). Materials per function is asymptoting toward molecular machines.

My trees are made with atmospheric sunlight, air, and water, with less than a kilogram of trace minerals from the soil. Nothing imported. Our machines will be almost that good, in an eyeblink of evolutionary time.

Minaturization. Nature thinks highly of it. We should consider it.

I help work on server sky, http://server-sky.com

Internet data centers are the fastest growing and most profitable use of electricity today (1000x markup!). It is far easier to ship bits than joules with microwaves - data transmission can easily tolerate 90 dB losses, while space-to-grid power beaming fails with more than 10 dB loss. Someday we will ship terawatts from space, but that will happen MUCH sooner if we can scale from megawatts to terawatts with something far more profitable and easier to deploy with existing technologies. What we propose may be too difficult in 2015, but every passing month adds technological capability. Whoever seizes first mover advantage will make the first trillion dollar fortune. As the spiders building a web at the bottom of Gary Larson's cartoon playground slide said, "if we pull this off, we eat like kings!"

Intel and Micron just announced their 3DXpoint memory technology, with the first prototype chips storing 128 gigabits in two layers of metal-to-metal grid crosspoints. They don't say, but the memory element between the wires is probably a tiny pillar of phase-change chalcogenide glass, which gets faster and cheaper and lower power as it scales down. It is also hellaciously radiation resistant, even better than the hafnium gate finfets underneath. Many details still secret, but if we assume they are using their current 22 nanometer wiring pitch technologies, then 128 Gbits of their double-stacked bit cells fit in a square die 6 mm on a side. They can keep stacking until the cows come home, and their wiring pitches are expected to shrink to 10 nanometers within four years. A 16 layer process may soon cram half a terabyte onto a similar 36 mm² chip.

Chips have grown denser and thinner. Almost a decade ago, I contributed technology to Hitachi's "powder RFID" project. Those chips were experimental, 50 micrometers square and 5 μm thick, with thousands of old-tech transistors. Today, far denser chips are routinely thinned to 20 μm in production; that saves some weight, but mostly it reduces the thermal resistance between the transistors and the heat sink. Thinning is basement lab technology. A 36 mm² chip, 20 μm thick, with an average density of 3000 kg/m², weighs 2 milligrams. At $10K/kg launch costs, it would cost 2 cents to put that chip into orbit. In zero gee and vacuum, you don't need packaging.

I don't know how much storage Google has worldwide, but in 2013 Randall Munroe estimated it at 15 exabytes - lets assume 50 eB a few years into the future, or 100 million hypothetical 2 milligram 500 GB chips. 200 kilograms of chips, launch cost $ 2M, to a place where the 2.7K heat sink is free, as is 1.37 GW/km² sunlight.

Power is more expensive - Google uses 700 MW, probably half of their power use is cooling, power conditioning, and support. Assume computation growth, and compensating architecture and efficiency improvements. 400 MW for their compute core in the future, half a large thermal power plant. With 15% efficient solar cells on a thin aluminum substrate, server sky will produce about one watt per gram, 200 W/m², so Google might need 2 km² weighing 400 metric tons, $4B to launch at $10,000/kg. Very expensive, but less than they spend on terrestrial data centers, and far less than launching far more watts of space solar power satellite, feeding an inefficient power transmission chain, to a data center spending less than half its power doing actual computing.

"Thinsats" will be 16 cm square, 5 grams and 5 peak watts each, in arrays of 8000 thinsat, 40 kg. 70 GHz, 25 dB atmospheric downlink loss from MEO to 45N, 60ms ping time. Hundreds to billions of arrays making 100 meter ground spots for point-to-point communication. Population doubling faster than yearly, light sail manuevering, and end-of-life recycling as ballast for next generation thinsats.

We have failed so far because we have turned Glaser's revolutionary 1960s ideas into ossified religious relics. We have new tools and new markets. I'm sure that if Glaser was alive today he would adapt to available technologies. His ideas were based on sober extrapolations of 1960s launch technologies - then the world went a different way.

We WILL import power from space someday, but if a torrent of data from space can displace exponentially increasing data center power consumption, and computation can help us use the rest of the world's power efficiently (the historical trend), we will create the time window, the wealth, and the launch capacity to develop and deploy power-to-ground SSPS. Real pioneers are clever and shave the angles, they don't build the Erie canal before they start making money. The way to skin an elephant is to practice on a mouse.

If a small, unfunded team can think all that up, imagine what 7 billion of us can do. I only expect a few million of us will actually go to libraries and labs and invent stuff, the rest will spend their lives on shopping malls and TV, and fear the world because that is what television teaches. Fear encourages short term thinking and unhealthy consumption; if the future is terrible, why save money and health, and spend time learning and inventing? But think about it - if the future is terrible, why do companies want your money, and why do they invest what you give them? What do they know about the future that you don't?

When we open the solar system, and the doom myths become impossible to sustain, the fear-consumption complex will collapse, and we will put a reduced flow of resources to much better use. I will not live long enough to see that, but it is fun to imagine the doomsters and hucksters thrown into the dustbin of history, while billions of newly empowered people fill the universe with life and intelligence, instead of their bellies with junk food and their minds with foolishness.

Yes, some people are doomed. Nature is cruel to the nonadaptive. Their places will be taken by those who choose to think, and thrive.

DoomIsUnsustainable (last edited 2015-08-20 19:08:34 by KeithLofstrom)