It is worth stressing that not everything is possible. In spite of the great potential diversity of possible extraterrestrial life, the writers of Star Trek, The X-Files, and the Alien movies (and some putative UFO abductees and their psychiatrists) have alas sometimes overshot the mark. An example is the screenwriterly propensity for portraying successful interspecies mating. (By this I don’t mean interspecies coupling, which happens every now and then on Earth, and with reckless abandon on Star Trek. There is the famous scene that escaped the censor’s notice in the 1960s between Captain Kirk and Queen Deela; there’s the love affair of Dr. Beverly Crusher and Ambassador Odan; and the dalliances carried on between the virile Commander William Riker and almost every alien in a skirt.) Of all the non-physics issues about which I received letters after my last book, this aspect of Star Trek seems to have provoked the most scorn—although I suspect that the human-alien hybrids on The X-Files generate less wrath among its viewers. On Earth it is well known among biologists and some farmers that copulation between species rarely produces viable offspring. The genetic code, while apparently infinitely malleable, is also quite sensitive. You might as well try running a Macintosh code on a Windows 95 system! Even species that are remarkably close in genetic makeup are biologically incompatible in matters of reproduction. And in the rare cases where offspring are viable—mules, for example—they themselves generally cannot reproduce.
Now, this is true of species that have coexisted on the same planet for perhaps millions of years, and have responded to similar sets of evolutionary imperatives, with genomes that are not markedly dissimilar. Imagine attempts at cross-breeding between two species that have evolved on separate planets. Even if the fundamental chemistry was the same—something not necessarily likely—it’s extremely difficult to believe that the product of mating, say, a Vulcan and a human being would produce anything as viable as Mr. Spock, any more than the coupling of a human and a chimpanzee would be likely to produce successful offspring. (My ordering should not be taken to suggest a correspondence to the Vulcan-human analogy.)
In any case, the fascinating new discoveries of the past few years have changed the way we think about the probabilities of life in the cosmos. Previously, the existence of planets outside our own solar system was pure speculation, and the range of conditions that might allow life to form and survive was thought to be far narrower than we now know it to be. At no time in this century has there been more reason for optimism about the possibilities of discovering extraterrestrial life, perhaps even intelligent extraterrestrial life, in our future.
For over 30 years, the standard estimate for the probability of the existence of extraterrestrial civilizations has been codified in what became known as the Drake equation—after the astronomer Frank Drake, who proposed it. In this equation, the number of intelligent civilizations in the galaxy is calculated as the product of the number of stars in the galaxy times several different probabilities expressed as fractions: the fraction of stars that are likely to have planetary systems; the fraction of these that are likely to have Earth-like planets; the fraction of such stars that are likely to be stable long enough for life to evolve; the fraction of these life-forms that are likely to evolve to achieve intelligence, and so on. In a sense, what this equation does is parameterize our ignorance, since each of the fundamental probabilities that goes into it is subject to debate. In this way, different groups have estimated the number of intelligent civilizations in our galaxy as ranging from millions to one. However, as time goes by and our knowledge increases, more reliable estimates for at least some of these factors have emerged.
Nevertheless, I have always felt that there is an inherent problem with this approach, and I recently had a discussion about it with Frank Drake himself, at the Naples conference on the search for extraterrestrial intelligence, which I mentioned in chapter 2. The point is that many of the individual probabilities whose product goes into the equation are small, and their product is even smaller. Thus, one goes from perhaps as many as 400 billion stars in our galaxy to perhaps only a handful of intelligent civilizations. Now, when probabilities get this small, they are sometimes difficult to estimate. The statistics of very rare events is quite subtle, and the most naive application of probabilities may not be the best way to approach this subject.
In the first place, whenever one considers a probability that results from the product of many different individual probabilities, the result has to be a small number, because each individual probability that goes into the product is less than 1, and the product of many numbers smaller than 1 is always very small. For example, the probability of any one particular event in your life taking place is, when viewed this way, almost zero. The probability that I woke up this morning in Geneva at 7:30 A.M. required first that I be on leave from my home institution, which in turn required that I be at that institution in the first place, which required that I chose physics as a career, and so on. More immediately, my waking up at 7:30 A.M. probably required that there be a small pond outside my window, in which a particular tadpole had become a frog that croaked at 7:29 A.M., and so forth. Though all these probabilities (and others too numerous to mention) were small, leading to an infinitesimally small probability that I would do exactly what I actually did, nevertheless I actually did it. Events with small probability happen all the time, because all events, when viewed in this way, have small probability.
By the way, this is one reason we have to be careful when someone tells us something like the following: “I had a dream the other night that my wife cried out to me as she fell down the stairs and broke her leg. A week later, she did trip and injure herself—isn’t that amazing? The probability that my dream would come true is so small that something fishy must have been at work here.” Well, to this notion the famous physicist Richard Feynman used to have an interesting rebuttal. He would sometimes exclaim, “You’ll never believe what happened to me this morning!” When you took the bait, he would answer, “Absolutely nothing special!” The point is that we tend to remember those events that stand out and forget those that don’t. An amazing coincidence is in any event amazing, but perhaps not as amazing as we might think.
There is a related problem one must confront here. If one considers the probability of many separate events occurring, one must also consider whether or not they are correlated—that is, whether or not they are truly independent. If they are correlated, simply multiplying individual probabilities will not give you the correct estimate, and the final probability may actually be much larger than one will predict if one makes this error. For example, the probability that I will utter an obscenity at any given instant may be small (although it is certainly not zero). The probability that I will hit my funny bone at any given instant is also small. However, the probability that I will hit my funny bone and then utter an obscenity is not equal to the product of the probabilities, since the probability of swearing at a given instant is correlated to the probability of hurting myself at a given instant. Similarly, the probability that a planet might survive meteoric and cometary impacts long enough for intelligent life to evolve may be small. And the probability that a solar system has a Jupiter-size planet in its outer reaches may also be small. But these two factors are not independent: The gravitational effect of Jupiter is believed to be important in deflecting many potentially lethal objects away from Earth’s orbit.
The modern parlance for these notions is “conditional probability.” Its expositors hold that we should not concern ourselves with “absolute probabilities,” which often have no relevance to things as they are, but with “conditional probabilities”—the chances that some event will occur when some set of previous conditions exists. However, we don’t always know what probabilities are conditional on other probabilities; as a result, things can get complicated when you try to estimate an exact probability that some specific complex set of events will occur—outside of those performed in controlled experiments in the laboratory.
A way has
been developed around this problem, based on the somewhat non-intuitive but extremely important notion that something that has a small absolute probability can nevertheless happen more frequently than any of the possible alternatives. As I’ve mentioned, every event that happens in the world can be viewed as having a vanishingly small probability if all the contingent factors are taken into account. What is therefore important is not the absolute probability but the relative probability. Given a wide variety of outcomes, what set of observations is more likely than others? If one set of possible outcomes has a raw probability of 1 in a million—well, that sounds pretty small. But if the other millions of sets of outcomes each have a probability closer to 1 in a billion, then the first set of outcomes is 1,000 times more likely to be observed in a single trial than is any other set of outcomes.
Of course, where so many possible outcomes are involved, what becomes operationally important is not one specific set of outcomes so much as whether the observed set is close to the one with maximum likelihood. An example should make this clearer. Say that I begin a series of coin flips and count the number of heads and tails. We all intuitively know that the maximum likelihood is that the number of heads will be approximately equal to the number of tails. However, we don’t expect the number of heads always to equal the number of tails. If I flip the coin 10 times, I may get 6 heads and 4 tails, or vice versa. As I flip the coin a larger and larger number of times, the number of different sets of possible outcomes continues to increase, and thus the probability of any specific set of outcomes (say, 499 heads and 501 tails out of 1,000 flips) gets smaller and smaller, precisely because there are more and more different possibilities that can occur. But despite the fact that the absolute probability of any specific combination decreases, the relative probability of getting very close to 50 percent heads and 50 percent tails gets higher and higher. By the time you have flipped a million times, the likelihood of deviating from this mean value by only 10 percent is 1,000 times smaller than the likelihood of lying within 1 percent of a 50-50 split! This is true despite the fact that the probability of getting 500,000 heads and 500,000 tails, the most probable outcome, is less than 1 in 1,000.
I can write down any specific tally of heads and tails that may result when I flip the coin a million times. It’s easy to calculate the probability of this specific set (say, HHTHTTTHTHT… ) occurring, since there’s only one way it can occur. Since with each flip the probability of a head, say, is .5 (that is, 50 percent), the probability of getting the sequence in question is (.5) × (.5) × (.5)… = (.5)1,000,000, which is, needless to say, a very small number.
So, since each specific sequence—even a T repeated a million times—has precisely the same probability as any other specific sequence, how come we never expect that at the end of the million flips we will have a million tails? Well, because there are many different ways of writing down a sequence of Hs and Ts that will end up with 500,000 Hs and 500,000 Ts, but there is one, and only one, way of writing down a sequence of a million Ts. It’s as simple as that.
What the technique of maximum likelihood does in this case is to find the characteristics of those types of sequences which have the maximum likelihood of occurring by comparing relative probabilities, without worrying about absolute probabilities, and also recognizing that any one particular sequence may be extremely rare. The method in this case would tell us that the sequence resulting in something close to 500,000 heads is much more probable than anything else, so that one of the possible sequences leading to this result is more likely to be observed than anything else, even though the probability of any one particular sequence occurring is extremely small.
Now, what is the point of all this when it comes to the possibility of extraterrestrial life in the universe? Well, what may be important to consider is not the absolute probability of any specific sequence of events leading to intelligent life, but rather the relative probability of some such sequence occurring compared to the probability that some sequence will occur which will not lead to life. It is the relative probability that is important. If we have learned anything over the past decade, it’s that life is more robust than we had imagined. I’m now more willing to assume that when you have organic material in the presence of some heat, some light, and some water, it’s difficult for life not to arise, even if the probability of its arising by any specific sequence of events is small. Instead of considering how probable it is that Earth-like conditions would obtain on any other planet, it might be more appropriate to ask, “What is the probability that organic materials will not by any route form self-replicating systems in several billion years on a given planet?”
I repeat that I have no idea of the answer to this question, and I emphasize that the answer lies outside my expertise. But it seems to me that, as in the coin example above, there could be many more routes to the evolution of life-forms than there may be to ensuring that a given solar system is devoid of life.
Once one thinks in these terms, focusing on the remarkably lucky specific series of circumstances that led to the evolution of intelligent life on Earth may be wide of the mark. If the likelihood of some type of life evolving on some system is greater than the chances of ensuring that no life at all arises, then—as remote as the probability of any particular sequence of events leading to life may be—we are more or less guaranteed that some such sequence will occur in most situations.
I am not suggesting that the Drake equation is flawed as it stands—it is not—nor that it needs on fundamental grounds to be replaced by the Krauss equation, even if that does have a nice ring to it. If we knew all the contingent factors leading to any type of life, we would be able to write down the probabilities exactly, and thus accurately determine the number of intelligent civilizations. And maybe one day we will be able to, since evolutionary biology is itself evolving by leaps and bounds. In advance of that knowledge, though, comparing relative probabilities may provide us with better insights.
Finally, there is an overriding factor suggesting that the formation of life—even intelligent life—may be possible or even common elsewhere. It is that we exist. This undoubted fact demonstrates that intelligent life can form under at least some subset of circumstances known to be present in the galaxy. Moreover, the lessons of natural history on Earth suggest that not only is life extremely robust, persisting even through mass extinctions, but also that the evolutionary routes leading to different complex organisms are numerous. In this regard, one should note with caution that while natural history tells us that life formed relatively quickly on Earth, it still took almost 4 billion years for intelligent life to evolve, and even then only by a series of historical accidents. This could well mean that life is common but intelligence isn’t. On the other hand, by the same argument given above, intelligent life might result from many different historical trajectories, and the one that produced us might be only one of many. Hard to know with a sample of only one!
In general, I suspect that since our own Sun is a rather ordinary star, and its place in the galaxy is unremarkable, and since nature repeats herself as often as the laws of physics and chemistry allow, it would be odd if life weren’t ubiquitous in the galaxy. It is just a matter of time—although perhaps on a cosmic timescale—not a matter of principle, I believe, before we discover our galactic cousins. I’ll go even further and say that I expect microscopic forms of life to be found elsewhere in our solar system within the next century. (Whether they will turn out to have a common origin with life on Earth is an open question.) The discovery of extraterrestrial intelligence, however, is doubtless much farther in the future, simply because of the near impossibility of round-trip travel to the stars, and also the difficulty of communicating across the vast abyss of space in the absence of agreed-upon forms of communication.
Look at it this way. Even without boats to travel across the Atlantic or Pacific, it is possible to send messages, or at least greetings, to civilizations on the other side of the world. Messages in a bottle have be
en discovered, for example, thousands of miles from their origin. Yet for about as long as it took European civilization to evolve to the point of transatlantic travel, there was no knowledge whatsoever of New World civilizations.
But unlike the first transatlantic explorers, who set sail with the intention of bringing back riches to their homelands, Earth’s first interstellar travelers will probably have no intention of returning. Like many a refugee, we will move out into the galaxy because we will have no other choice. The laws of physics, not the laws of mankind, will require us to leave.
CHAPTER EIGHT
THE RESTAURANT AT THE END OF THE UNIVERSE
The choice is: the universe… or nothing.
—H. G. Wells
With the approach of the millennium, now is an appropriate time to join the crowd and proclaim the inevitable: The world is going to end. When—and, perhaps more important, how—are issues that are not quite so clear.
On the whole, I think Doomsday has gotten bad press. I will argue here that it holds great potential for the human race. With typical astronomical precision, we can pinpoint an upper limit for human existence on Earth at about 5 billion years from now, give or take 500 million years. So there’s still time to get your broker on the phone and sell your stocks, and don’t give up that reservation in Paris for Christmas 1999. Still, in the language of logicians, this upper limit, which marks the whole planet’s termination, is a sufficient but not necessary date for our demise. We could easily perish long before, in a global Armageddon, or because of some new efficient virus, or because of an astronomical catastrophe such as a large meteor impact. Or, of course, aliens could decide to annihilate us.
Beyond Star Trek Page 7