Printed: 2020-06-03

Institute for Ethics and Emerging Technologies

IEET Link:

The Myth of the Starship

Charlie Stross

Charlie's Diary

November 29, 2009

As starships do not in fact exist, no starships were harmed in the production of this essay. Also, this is just words. If they upset you, go lie down in a dark room for half an hour then drink a glass of water; you’ll feel better.

Actually, I tell a lie. There are five starships that we know of; Pioneer 10, Pioneer 11, Voyager 1, Voyager 2, and New Horizons. But they’re a far cry from the gleaming interstellar transports of science fiction. New Horizons is the most recent of them. Launched in late 2006, it is the fastest human-launched vehicle so far. It raced past Lunar orbit within nine hours of take-off: nevertheless, it will take around 10 years to reach Pluto (its proximate target — for a three-hour flyby). It weighs around 478 Kg, and is currently travelling outwards from the sun at around 17km/sec — about fifty times as fast as a rifle bullet.

We are 4.37 light years, or 140 million light-seconds, from Alpha Centauri, give or take. One light second is 300,000 km; it takes New Horizons about five hours to travel one light second. So: in very roughly 30 million days, or on the order of 300,000 years (if it was going in the right direction, which it isn’t), New Horizons could reach Alpha Centauri.

And that’s the best we’ve done to date, admittedly without really trying ...

This is not an essay about whether we could do better if we tried. I’ve written about the problem of space colonization before. Rather, what intrigues me is the possibility that the entire conceptual framework of the starship is a dangerously misleading dead-end, and that what we need is a new framework for thinking about interstellar travel.

The very word “starship” is a concatenation of two other words — star, and ship. The first is pretty harmless; it merely defines the scale factor we’re talking about, as opposed to interplanetary ship, or moon ship, or Atlantic-crossing steam ship. But the second word comes with a whole freightload of unwanted baggage, and I’m of a mind that serious futurists or SF writers might want to think about ditching it completely and looking for something new.

The astute reader will have spotted the link to the Apollo Program above. We have actually built and flown Moon ships ... but we didn’t call them ships, and they didn’t much look like one of these. There are several reasons why not. First and foremost are the scale factors involved, scale factors in both distance and (because you need to cross that distance) time, hence velocity (the change of distance over time). Kinetic energy is the killer; the kinetic energy of a moving body is proportional to the square of its velocity. Want to go faster? You need to throw bucketloads more energy at your vehicle, or make it much lighter. The Apollo spacecraft (or New Horizons) travel roughly three orders of magnitude faster than a regular sea-going ship, and they also had to haul along the reaction mass to throw out the back to get them up to speed (for they’re all powered by rockets, relying on Newton’s Third Law to make things balance out).

But there’s a more subtle difference. We have a long tradition of nautical baggage. Seafaring ships of the great age of exploration were largely wooden, and — with the aid of their human crew — self-repairing; subject to the availability of raw materials, there wasn’t much aboard a 16th or 17th century sailing ship that couldn’t be made on board. Aside from carpentry, the inhabitants of even a relatively small port could make the necessities to keep a ship at sea on a voyage of years — a smithy, a pottery, a glass-blower,  weavers of sailcloth and makers of hardtack. Shipbuilding was by no means easy (it was an economic activity born on the backs of the large numbers of peasant farmers and fisherfolk it took to provide the surplus to feed the workers in the shipyards) but it wasn’t anything like the Apollo project, which sucked up the labour of a third of a million skilled engineers and technicians for a decade. The word ship therefore comes freighted with connotations of autonomy and sustainability that are inappropriate to space travel as we know it today. And one of the most perfidious, misleading, damning, unconscious associations of the word “ship” is the word “destination”.

As I’ve said before, the trouble with going into space is that there’s no “there” there when you get to the other end of your voyage. All you get is rocks, sunlight, and if you’re lucky some slush (water optional: could equally well be methane or cyanide) to season your gravity wells. So if you want to do anything at the other end — anything beyond looking around real good — you need to bring the minimal seeds of the infrastructure with which to build and maintain your biosphere (if you’re travelling with an entourage of meat puppets) or your mechanosphere (if you’re going the Eric Drexler/Hans Moravec/downloading-or-AI route).

Which is why I was asking questions like this and this and this about a month ago. I was feeling my way towards this critical question: which is, “how simple can you make a minimal self-maintaining interstellar transport system”?

To a first approximation, the best answer I can come up with is “not very”. We can probably make it mechanically simple, rugged, and lightweight if we can do mature machine-phase diamond-substrate nanotechnology, and if we can figure out how to do one of mind uploading or artificial general intelligence

Note that I say mechanically simple — there’s a monumental raft of complexity wrapped up in the idea of using a Starwisp for establishing interstellar transportation, but it’s informational complexity rather than straightforward mechanical complexity. We may use such a system to sidestep the need for learning how to build self-sustaining biospheres and interstellar playpens for bored hominids, and how to equip a group of said hominids with the wherewithal to keep such a mobile playpen from degrading catastrophically, but we face the corresponding monumental challenge of solving the hard-AI problem and developing molecular manufacturing far beyond the flexibility and scope of today’s nanotechnology applications.

Such an interstellar capability isn’t going to look much like a “ship”. It’s going to look more like a DVD balanced on a microwave beam, or a can of beans hanging below a light sail energized by lasers powered by huge orbiting solar power stations. There won’t be any biological agencies aboard: just AGIs or something equivalent ported out of a fleshbody’s cranium. No hands, only nanotech assemblers. And after a voyage of decades or centuries it’s going to have to stop — somehow braking at the other end — then spend more decades farming rocks, slush and sunlight to build ever-bigger physical structures until it can build the equipment with which to phone home

If anything, it’s going to resemble a seed pod for a different kind of life, and on arrival it’s going to hatch and grow into a tree, or a forest, or a manufacturing-industrial complex. Finally, long after arrival, it might have sufficient resources to divert from homeostasis and growth to construct a biosphere, open communications with home, and prepare to download digitized colonists — if the whole uploading concept doesn’t prove to be chimerical, and if there’s something to be done with the serialized primate core-dumps at the other end.

Note that I’m fairly optimistic about mature diamond-phase nanotechnology (or some cognate thereof). The economic benefits of getting it are huge, and there are no obvious lacunae on the technology road map — unlike, say, fusion or manned interplanetary space travel. I’m less optimistic about mind uploading, because in neuroscience we are just about at the stage of beginning to figure out how ignorant we are. And I’m pessimistic about AGI, because I don’t think we stand a hope in hell of working out how to design an artificial general intelligence until we know, at least in outline, what human general intelligence is. (And we don’t.) But I suspect some combination of these technologies will show up sooner or later — barring resource-depletion crashes and/or habitable-biosphere-envelope departures on a planetary scale — and once you’ve got two out of the three, a starwisp-driven expansion starts to look feasible (if energetically expensive).

The alternative approach uses generation ships. That’s what I was talking about earlier. Which doesn’t sound much easier, if any, when you contemplate how much we don’t know about environmental engineering and biology. It also begs some huge sociological — human — questions, and is unlikely to be planned as such. I expect it’d only happen in the wake of our development of stable, safe space habitats — itself a huge obstacle (Just look at the history of our first modular space station if you want to get an idea of the problems surrounding life in space) — and if the energy costs required to launch a starwisp are high (56Gw for about a month to punch a 1Kg payload up to 10% of lightspeed; a microwave lens 560Km in diameter to focus the beam), those of a generation ship are going to be astronomically higher.

... Which leads me to believe that we’d be more realistic if we just ditched the word “ship” entirely from discussions of interstellar travel.

What we need to contemplate are the requirements for of an interstellar transportation system.

Such a system needs to provide not only a mechanism for sending a self-replicating technosphere across interstellar distances; it needs to be able to produce a habitable space at the destination, and provide a return option (for data, if nothing else).

Waving a magic wand of the variety I discounted in my earlier space colonization essay, even if we postulate that mind uploading is both possible and the way forward, interstellar travel still won’t be cheap. Direct communication via modulated laser looks feasible at extrasolar distances, on a reasonable power budget, given adequate pointing accuracy — but the study linked to above has a bit rate of only 0.5 kb/s. Given their 15w peak power output, they’re talking about 0.03 joules/bit across 1000AU distance; or 1.9 joules/bit/light-year.

Even Hans Moravec’s estimate of the computational complexity of the human brain (cited here only as a starting point for discussion of mind uploading: I think he’s erring on the optimistic side by between three and six orders of magnitude) suggests you’d need many years to transmit a compressed uploaded mind at that bit rate! But we know we can transmit data much faster using lasers over optical fibre. If we can push the bit rate towards 1Tb/sec, transmitting a map of a brain with 10 to the 11th neurons and glial cells, each with on average 10 to the 4th interactions with neighbouring cells, and, say, a spare 30 bits/connection (to summarize its properties),  for a total of 10 to the 18th bits of data per upload, we can squirt that map at the stars in about ten days. We’ll need a powerful laser unless we want the error rate to climb. Using the figure of 1.9 joules/bit/light-year, to send an uploaded mind to alpha centauri is going to cost on the order of 10 to the 19th joules over ten days: an unfeasibly large amount of energy. (Hint: we’re talking the equivalent of a Hiroshima-sized nuke every second, for a million seconds).

However, that figure corresponds to the sort of bit rate we can envisage achieving more or less now. Improve the data-to-power ratio by, say, a millionfold and things begin to look feasible (if not exactly cheap). In particular, it’s a lot cheaper than travelling in person (and anyway, we’ve got the honking great big solar power station/maser grid we built to launch the starwisp in the first place; might as well put it to use). Finally, if we can do the upload thing well enough to provide the brains to run our starwisp in the first place, we can probably make do without building a biosphere at the other end. In other words: frail fleshbodies need not apply.

(There’s an alternative to shipping around uploads via laser that merits investigation: if we can do uploading, and if we can make memory diamond — which would seem to be a reasonable expectation of a mature machine-phase nanotechnology — then the 80g payload of the reference starwisp ought to be sufficient to carry about 2 x 10 to the 24th bits, which corresponds to 20,000 stored uploads per “passenger ship”. This might well be energetically cheaper than using a laser to transmit uploads, giving us an unexpected long-haul corollary to Tanenbaum’s law.)

So, to summarize: yes, I think human interstellar exploration (and yes, maybe even colonization) might be possible, after a fashion. But to get there, we’re going to have to master at least two entire technological fields that don’t yet exist, even before we start trying to blast compact disc sized machines up to relativistic velocities. And that’s without considering the difficulty of how to cram an industrial infrastructure capable of building more of itself, of a machine capable of surviving in deep space — the equivalent of those 300,000 NASA technicians and engineers — into the aforementioned CD-sized machine ...

If we succeed in doing it, it’s going to look nothing like the Starship Enterprise. Or even New Horizons. The whole reference frame we instinctively assume when we hear the word “ship” is just so wrong it’s beyond wrong-ness: it’s on a par with Baron Munchausen’s lunar exploits as seen in light of the Apollo Program. We need a new handle for discussing and analyzing such a venture. And the sooner we consign the “-ship” suffix to the dustbin of failed ideas, the better.


Contact: Executive Director, Dr. James J. Hughes,
IEET, 35 Harbor Point Blvd, #404, Boston, MA 02125-3242 USA
phone: 860-428-1837