A new study that ran the Drake Equation with a proper sampling of uncertainty (something no one had done before, oddly) has found there is credibly a 38% chance humans are the only civilization in the entire known universe; and a 52% chance we are the only civilization in our own galaxy. Not ever, to be clear. But, alive, within a detectable window for us. In other words, there could be many other dead civilizations we missed the opportunity to see, and many others whose light can’t have reached us yet. But excluding those, the most likely probabilities on current scientific knowledge and established uncertainties, are that there is only a 62% chance another civilization exists anywhere in the universe whose signal could be reaching us; and only a 48% chance of there being one in our own galaxy.

This work is interesting and worth discussing. It has implications for methodology as well as naturalist philosophy and Christian apologetics. Here’s a breakdown…

The Basic Idea

The paper, currently being considered by the Proceedings of the Royal Society of London A, is “Dissolving the Fermi Paradox” by Anders Sandberg, Eric Drexler and Toby Ord (of the “Future of Humanity Institute” at Oxford University). Peer review may lead to updates or revisions in the final published version (so keep your eye out for that), but their paper is solid, straightforward, and contains all the correct caveats, so I doubt it will be rejected.

What they did was take the Drake Equation, which allows boundary conditions to be entered for all the requirements for there to be a detectable civilization currently cohabiting the galaxy with us, and then calculate what the probability is that we are alone in the galaxy given our uncertainty about the variables. And then repeated the procedure for the observable universe. In other words, given that we don’t know where each parameter falls between two values, we can randomize where it does fall within the range of possibilities, and average out the resulting probabilities, to get a “so far as we know” probability of where that parameter is.

There are other forms of the equation than the standard one they use (see for example Wikipedia’s sample of some). But as they note, their results generalize to all of them. Unless some factor one wishes to add plausibly varies by more than an order of magnitude, and odds are none do—and even if they did, most would in consequence increase the probability we are alone in the visible universe.

I have to be additionally clear, though. The Drake Equation (hence their results) only pertain to living civilizations. You’d need a different equation to calculate how many civilizations there have been in this galaxy or the known universe up to now (and yet another to calculate how many there will have been by some point in future time). And the probability that there are one or more dead civilizations in each zone, is necessarily higher than that there is a still-active one. Although “still active” is not quite the right term here either: Drake gives us the number of civilizations whose signals or activity we should be able to see (leading to some then adding a factor to the equation regarding whether or how long civilizations send signals or leave evidence at all); and it’s possible a civilization died out after sending any such signals or light evincing such evidence, but the difference there is just a time lag, so it doesn’t affect the “how many we should see by now” factor.

Which is what generates the Fermi Paradox this paper is actually about: we don’t see any. Why? There are really only two possible answers: advanced civilizations are invisible; or there are none to see. This paper supports the latter conclusion. And it doesn’t come to that conclusion arbitrarily. It looks at what would have to be the case for there to be any to see (if they were visible), and finds that we have no evidence those conditions have been met, and what we do know strongly suggests in fact they haven’t been. So we don’t need to posit that they are invisible.

Invisible Civilizations

Of course by “civilizations” here we mean not merely integrated networks of cities, but actually EM-radiating societies. So earth only counted as a “civilization” in this equation when we started beaming radio signals, some of which inevitably leak into space. Thousands of years from now, some distant world as many lightyears away (if it has sensitive enough equipment) could detect our signals. Ergo the other way around. Hence SETI. (Nevertheless I will use “civilization” more broadly, unless I am referring to the Drake.)

This becomes all the more the case if we suppose we will inevitably begin colonizing the galaxy. The galaxy has been around for billions of years, yet it would take mere millions of years to colonize most of it, so there has been a thousand times more time than is needed for any species that arose ahead of us to already be occupying the whole galaxy. Unless such species are so rare, none arose millions of years ahead of us, anywhere in the whole galaxy. But for that to be the case, such civilizations must be so extremely rare there might not be any at all (other than us). Hence this new study’s results.

Some don’t like this conclusion. So like Christian apologists, they try making up excuses for why the evidence doesn’t fit expectations. Though not always implausibly.

Maybe civilizations don’t colonize galaxies. But we have to suppose this is so extremely commonly the case, that we must expect virtually 100% of the time that no civilization will. Not even we will. Even if our civilization lasts forever. Maybe we will be content to just sit around in one place, even after the sun burns out. Though this is possible, it really doesn’t look like anything near a certainty. To expect that every civilization, even hundreds or thousands of them, every single one, would make this same weird decision, requires a faith implausible (or so it would seem…we’ll revisit this assumption in a moment). And even then, an EM-radiating civilization sitting in one place should still be detectable eventually (another assumption we’ll revisit in a moment). Though we now know our own signals barely escape our heliosphere, it is quantum mechanically inevitable that an instrument sensitive enough will always pick up any signal (up to certain limits); so detection is simply a function of instrument sensitivity (or being too far away). So maybe we don’t have sensitive enough instruments yet?

Or maybe civilizations all reach a point when they don’t radiate. This is actually a more credible hypothesis. One can show that certain forms of progress in technology are inevitable, that they have a virtually 100% certainty over, say, a million year timeline—and remember, our civilization has only been around for a few thousand years, and radiating for barely a hundred, so when you stop to think where we will inevitably be in a million years, you will get my meaning. Even if we experience several more dark ages in that span, the outcome of a super-advanced civilization is simply unstoppable—but for extremely rare accidents, which will barely put a dent in the number of civilizations that get there (civilizations are a lot harder to kill than people think).

We can show, for instance, that, averaged over time, it’s reasonable to conclude civilizations always inevitably advance in EROI: their ability to exploit and leverage energy. As I’ve discussed already, we’ve gone from human and animal energy (EROI in the low single digits), to environmental energy (water, wind, and biofuels: EROI in the high single digits), to fossil energy (coal in the 1800’s got our EROI above 10; fossil fuels probably max out now somewhere around 30), to nuclear (EROI above 100). When we get to antimatter and other forms of advanced power exploits our EROI will skyrocket. But notice it took us not even 10,000 years (a pittance of time in a galaxy over ten billion years old) to get from an EROI of 1 to 100. It’s reasonable to assume all civilizations inevitably make that same curve. Even if they take longer to do it—because even at “ten times longer” that’s still just 0.1 million years, a mere drop in the bucket of galactic time.

The only way for this not to be the case is for the civilization to get wiped out (meaning, the whole species extinct), which is extremely improbable (despite what doomsdayers keep saying); or for the civilization to choose to revert to lower EROI’s (like the planet in the  Star Trek TNG episode “Devil’s Due”), which must also be extremely improbable—as it requires a coordinated choice among (likely by then) billions of individuals to all revert to less effective and more expensive ways of accomplishing their goals. And never going back. Like that civilization that just sits around even after its own sun burns out. Civs that do that, are just not going to be that frequent.

Sandberg et al. decry the idea of positing “all civs go dark” to resolve the Fermi Paradox—instead of admitting (as they would prefer) that it’s resolved by there being no other civilizations—based on the notion that it seems incredible to suppose that this outcome is so likely as to always happen, even across hundreds or even thousands of civilizations. But though that’s a reasonable objection (as I’ve just noted), it might carry less weight in this case than they think. Would it be reasonable to suppose post-space-age school children (e.g. American children today) will still be using wax tablets to do their homework instead of digital ones? Not really. That virtually 100% of all civilizations will make that switch, is actually what we should expect.

And here is where “all civs go dark” might be just as certain an outcome. If civilizations always advance in EROI, they will certainly advance in efficiency, and thus their radiation will get less and less dirty (comms will more often be by laser, for example, than just throwing radio waves everywhere; consider how we are already transitioning from broadcast television to cable, as another example). That’s pretty much 100% certain. And would they want to draw attention to themselves by deliberately producing directed high energy signals? Arguably no. It seems a waste of resources, all just to advertise to potential invaders where you are. The few who try it even now, aren’t getting signals anywhere more than to a microscopic fraction of the galaxy.

But beyond all that, any fashion for it that may arise will soon wane…when we lose interest in this universe altogether. Remember, we are thinking about where we will be as a civilization a million years from now. And thus where all civilizations likely will as well. By then we will certainly have the ability to ditch this messy, brutal, murderous, disobedient, and uncooperative universe and go live in a far more hospitable one of our own making—one that will not be brutal, murderous, or disobedient, but that will conform exactly to our wishes, resolving all EROI limitations altogether. Paradise, in other words. We will live in simulated universes of our own design. Why would anyone prefer anywhere else? The odds that they will—and will continue to for thousands of years—are surely dismally low (though contrary to Bostrom, odds are, we aren’t in one of those…yet).

If you are going to transition your civilization to a simulated universe, you need to park it somewhere where it will last forever, and that means far away from all the shit that explodes and blasts you with deadly particle beams and the endless mess of deadly rocks and dust that will inevitably crash into you and fuck up your shit. Which leaves only one place to park: the intergalactic void. You will be starved for energy out there, but that’s not a problem, because in a simulated universe you can set your clock at any speed. You can let your servers tick a second of time over every few billion years in realspace, using the intervening period to soak up all the photons and energetic particles there are and use them to power the system. And radical efficiency here will be a must—something far beyond current technologies of efficiency, of course. Which means no such simverse-server-array will be radiating anything. Not even heat; at least beyond trivially (or observably, if they found a realm of energy transfer that doesn’t even interact with normal matter).

Frankly, IMO, this all follows with near 100% certainty. Nearly all civs will do this—because they can, and once they can, why on earth wouldn’t they? All who do will leave the galaxy, not colonize it. And all who do that will minimize all radiating signals. So actually, odds are, all civilizations are invisible. They went dark. They are just coasting in between galaxies, impossible ever to detect. Which doesn’t help us resolve how many there are.

Rare Earths

Sandberg et al. notice that of all the parameters of the standard Drake Equation, we actually know them all to within a reasonable margin, except the probability of abiogenesis. All the terms before that, science has constrained pretty well by now. And all the terms after that, are constrained by basic principles of probability, in particular, the rule that, absent evidence to the contrary, “you are always more likely to be typical than exceptional,” which is a tautology, but an important one. When you look at the final terms in the Drake Equation, we have an observation datapoint to work with, to which that principle would apply: our planet’s history.

How long will a Drake civilization last? Probably forever, or at least trillions of years (contrary to the doomsday crowd, it is, again, extremely hard to kill one), in any case well longer than the universe itself has even been around by this point. How many intelligent species will develop Drake civilizations? Probably all of them, or near enough. Once we had an intelligent species, we got a Drake civilization in a mere quarter million years, a microfraction of time. Life had been evolving for over three billion years, and yet as soon as an intelligent species arose, it almost immediately developed a Drake civ. “We are more likely typical than exceptional,” therefore this is almost certainty within an order of magnitude of how long it takes on average. For instance, to suppose it was ten times faster than average, is to suppose we were bizarrely—which means very improbably—lucky.

Almost the same could be said for the evolution of intelligence: it took only 3.5 billion years, and this can’t to any credible probability be even as much as ten times faster than average. And it’s vastly less probable still that we got lucky on both measures at the same time (intelligence and civilization). Although here, maybe not:

• It’s possible most life dies out well before even reaching the average timeframe for evolving intelligence. In other words, all biospheres would eventually generate intelligence…if all biospheres lasted forever. With evolution you are basically running a semi-random search routine for intelligence. So it’s statistically inevitable to find it. But since biospheres don’t last forever, most might go extinct before finding it. So the rate at which planets harboring life will produce intelligence is partly a function of the rate at which planets harboring life die off before getting that result.
• Therefore, if the average timeframe for life developing intelligence is longer than planets usually even remain habitable, then selection bias will ensure our timeframe is far faster than average. Because that will always be the case for every intelligent species and civilization observing itself (per the weak anthropic principle). The Drake Equation would reflect this with a low frequency of inhabited planets producing intelligence—since (under this supposition) most would burn out before they get there, while the rest will fall within the time-frame of habitability. Which for the earth is about 6 to 8 billion years, but can be for some planets many times longer.
• Civilization clearly does not meet this condition. It arose far too quickly, once intelligence existed, for selection bias to explain that. It is therefore improbable that the pace of its development deviated very far from average. It’s very unlikely to have been even as much as ten times faster than average. But the evolution of intelligence does meet this condition: it took a timescale here, on earth, pretty close to the limit (roughly half the lifetime of the planet, thus right smack in the middle of where it had to occur). Which looks more like selection bias, rather than indicating its actual frequency.

So the highest probability is that we are close to the hump of the bell curve on the transition from intelligence to a Drake civilization: that rate is probably pretty normal; but we might not be close for the previous stage, the transition from life to intelligent life: that pace might be pretty unusual (we can’t tell with just the one data point). So it is reasonable to conclude that it’s usually the case that once intelligent life exists, if nothing stops it, it almost always becomes a Drake civilization in under a million years. And there aren’t a lot of things likely to stop it that can happen in so short a cosmic time-frame. But the evolution of intelligent life probably takes at least billions (as here) if not on average tens of billions of years (averaged across every instance there may be). Maybe more.

We could try to argue for the contrary conclusion. Intelligent life on earth certainly required multicellularity; and this is probably universally true (we can’t honestly think of any other way nature alone could do it). But look at the clock: as soon as multicellularity was achieved, after running a sampling program for over three billion years, it then produced intelligence in only half a billion years. That’s extraordinary if it isn’t ordinary. So odds are, once that innovation arises, intelligence is rapidly inevitable, occurring on average in under a billion years. Anything else would make us very unusual, and without any obvious selection bias to otherwise explain it, unusual is improbable. So maybe multicellularity is what earth life discovered unusually early? Selection bias could still explain that; as then no intelligent beings would ever observe being anywhere else but on such lucky planets.

But time isn’t the only factor. It may be that life arising is so easy, that it arises usually in places where it can’t evolve into intelligence. For instance, if life originated on Titan, odds are the extremely low energy flows there would never be significant enough even to power the evolution of an intelligent species. You need a more energetic location. Like the earth.

This is essentially the thesis of Ward and Brownlee, who keep getting cited by Christian apologists as proving life is too rare to have arisen by chance, even though they never argued such a thing and in fact argue quite the contrary. They argue that probably life originates quite often in the cosmos, maybe hundreds or thousands of times per galaxy, but the conditions where it could arise are so variable that most worlds would never be hospitable to what they call “complex” life. Consequently, complex life would never evolve in most cases, even as life itself is found all over. And intelligent life is a very complex form of life. It requires a lot of energy to run, and it requires environments that aren’t too radically varying or extreme. Think of the bacteria that survive in Antarctic ice or deep in the crushing depths of the earth’s crust: life can thrive there, but could anything complex even move, much less survive in such conditions? Probably not.

As I’ve noted before:

[Ward & Brownlee] argue we may be the only intelligent species in this “quadrant of the galaxy” (pp. xxiv & 283), and are therefore only “virtually unique,” not literally unique. They do speculate whether we may be alone in this galaxy altogether, or the visible (not the entire) universe, but their actual thesis only extends to our galactic quadrant. There are over a trillion known galaxies. Therefore, their idea of ‘rare’ still allows for billions of planetary civilizations.

So I would agree, given the uncertainty of both the time problem and the environment problem, the frequency of life that becomes intelligent must surely be low. But the variances I don’t think could be too wild. Could it even differ by an order of magnitude? Are there ten times more places in the cosmos life can arise than can evolve the complexity of intelligence? Did intelligence arise on earth ten times faster than on average it would? Either seems a bit of a stretch to imagine. Particularly as life is far more likely to evolve in places where such room to evolve is also available. Nevertheless, Sandberg et al. look at the variance in published estimates of this and find (and thus employ) a variation of four orders of magnitude. That seems a tad absurd to me. But it’s reasonable a fortiori. They are, after all, averaging professional scientific estimates. They likewise find a variability of two orders of magnitude in the probability of intelligent life building a Drake civilization, which I also consider absurdly pessimistic—because it violates the basic rule that, absent evidence to the contrary, we are more likely typical than exceptional, and estimates that are that variable entail we are absurdly exceptional. But again, I’ll take it a fortiori.

The Sandberg team ends up, though, with the largest variability in the singular parameter of the probability of life arising anywhere at all. This they find varying in the literature by as much as 200 orders of magnitude! (p. 9) Even their next most widely varying estimate, lifespan of Drake civilizations, ranges over a space of only 9 orders of magnitude. Every other parameter varies by no more than 1 to 4 orders of magnitude in the literature. So the most important observation here is the huge uncertainty in the probability of abiogenesis. Which is what Christians love to focus on.

All this, Sandberg and gang conclude, once you plug all these uncertainties in, gives us a Bayesian prior probability that there is at least one other space-faring civilization besides ours that we’d currently be able to observe (if we could observe every one there were) of only 62% for the observable universe. That’s over a radius of about forty-six billion lightyears (objects that emitted the light now reaching us thirteen billion years ago having moved a considerable distance further since then), encompassing over a trillion galaxies. And it’s just 48% for our own galaxy. In other words, roughly speaking, for all we honestly know, there’s only a 50/50 chance there is even one other civilization like ours in our galaxy (that we could see). Extending to the whole known cosmos, it’s not much better. We could very easily be the only one in the observable universe (all others having died out ages ago, or arisen before any light from them could yet reach us).

Of course, the actual universe is vastly larger than the observable part of it. In fact the universe might be infinitely large; but observations fix its minimum size at 250 times larger than the observable part by volume (several times the observed radius). And that’s just the minimum. The currently most credible cosmological theory—the inflationary Big Bang model—entails the actual cosmos has a radius 10^23 times greater than observed. That’s a 10 followed by 23 zeroes. An absurdly vast size. So the probability of there being other civilizations like ours does start to approach 100% once you start counting regions of the cosmos beyond our visible horizon. And even counting just the observable universe it’s close to that, if we talk about all civilizations. When you remember the Drake equation is only counting observable civilizations (not long dead ones or relevantly new ones; which may actually be most civilizations), the probability of there having been many civilizations in the universe up to now (and not just us) is certainly well near 100%—we just won’t have received any light from them to tell by, it either all having long passed us, or not gotten to us yet.

You might still ask why the incongruity between the cumulative probabilities. If there is a roughly 50/50 chance of a civilization per galaxy, why is there roughly only a 2 in 3 chance of a civilization in any visible galaxy? The answer has to do with the uncertainties (shown on their Figure 2, partially reproduced above), which they explain (on pp. 10-11):

[T]he mean for N [the number of observable civilizations] is very optimistic, at 27 million, but the median is now only 0.32—less than one civilization per galaxy like our own. The probability of N < 1 is now 52% … Most markedly, the very uncertain life formation rate produces a heavy left tail, giving a nearly 38% credence that N < 10 [to the power of] -10, making us alone in the observable universe.

In other words, they are giving the median value rather than the mean value. So when we factor in how uncertain we are—particularly with respect to the frequency of abiogenesis—we end up with a high probability the cosmic civilization density is extremely low. For instance, if we put error bars around these numbers, they would be extremely wide, and hugely overlap, and probability consistency would be maintained at the margins, not the medians. This is a “for all we know” probability based on highly uncertain variables.

They then complete the Bayesian equation by using the Fermi observation as evidence that can now update our priors. Although there, again, margins of error are wide. So they only end up with this evidence raising “our 52% credence for being alone in the galaxy to somewhere between 53% and 99.6%” and “our 38% credence for being alone in the observable universe to somewhere between 39% and 85%” (p. 13). So at the margins, not a huge change. The evidence of no observable civilizations is just highly consistent with their prior. Odds are, there aren’t many to observe. [Update: A new paper has come out supporting their conclusion from a different angle, which is also worth considering: Snyder-Beattie et al., “The Timing of Evolutionary Transitions Suggests Intelligent Life is Rare,” Astrobiology (Mar 2021): 265-78.]

There is a serious significance to all this that actually reverses the Christian’s desired argument…

The Abiogenesis Nonproblem

Notably, the bottom of the scale of probabilities assigned by Sandberg at al. for the probability of a life-forming event on a given planet is 1 in 10^200. The top of the scale is 1 (meaning, life forms on every habitable “planet”). That’s about as wide open an error margin as you can assign here! Both are surely false. As in, their top is way too probable; their bottom, way too improbable. Which well encompasses the full range of our uncertainty.

Their top margin reflects current scientific hopes. The quest for extraterrestrial life in our solar system, involving actual funded missions to Mars (and there’s talk of hope for finding life on Titan, Europa, and elsewhere), presumes this frequency is indeed fairly near 1—such that on every moon-or-planet it could have happened, we might expect to find life at least got started even if it didn’t last. Their bottom is based on the random assembly of a folded protein in modern biology: the number of times per roughly 14 billion years that we would see “the folding of a moderate-size protein” anywhere in the universe without the biological mechanisms we know now accomplish this, i.e. for this to occur just “at random,” and presumably thereby produce a self-replicating prion, on which evolution by natural selection then operates. That’s as absurdly low a probability as abiogenesis could ever have.

All actual peer reviewed theories of self-replicator assembly are nowhere near that random, nor require self-replicators of such complexity as a “moderate-size protein.” They involve known and documented processes that order that assembly, from the spontaneous assembly of self-replicating RNA on the surfaces of clays, to the autocatalytic assembly of PNA molecules evolving into RNA molecules evolving into DNA molecules (and thence into the protein-folding cellular machinery found everywhere on earth today). And self-replicating PNA and RNA molecules are already known, that are far simpler in construction than a “moderate-size protein” (and countless more may be possible).

Reasonable estimates place even the random assembly of the simplest known self-replicating peptide molecule at around 1 in 10^41, according to the reasonable calculation of Johnjoe McFadden, in his otherwise ridiculous Quantum Evolution (2000: 98). That peptide was first described by David Lee in “A Self-replicating Peptide,” Nature 382 (1996): 525-28. Involving just 32 amino acids, with a simpler and more variable backbone than either RNA or DNA require, this molecule is far simpler than pretty much any protein known, much less of “moderate-size.” And McFadden’s calculation is conservative, because it operates on assumptions (such as that there are only or always 20 amino acids to select from in a random assembly, and that none can be substituted for each other) that, when replaced with more accurate science, make the random assembly more probable.