The Aliens Are Coming!

by Ben Miller

  THE GIRL WHO TALKED TO DOLPHINS23

  The interest from Frank Drake, Carl Sagan, and others helped Lilly to secure NASA funding for his Communication Research Institute, and in 1965 he conducted one of the all-time strangest scientific experiments, when he agreed that a twenty-two-year-old researcher, Margaret Howe, should be isolated for ten weeks with a young male bottlenose dolphin called Peter. Howe had been attempting to teach Peter to speak by repeating English letters, numbers, and words and trying to get Peter to “say” them back to her, much as a mother might teach a child to speak. Howe became convinced that she would make more progress if she was in constant contact with Peter, and persuaded Lilly to allow her and the dolphin to cohabit.

  The upstairs of the house was plastered and flooded so that Howe and Peter could share the same living space, with the knee-deep water being just shallow enough for Howe to wade through, and just deep enough for Peter to swim in. Howe slept on a foam mattress in the middle of one of the pools, worked at a desk suspended from the ceiling, and ate tinned food to minimize contact with outsiders. The days were run to a strict timetable, with Lilly giving precise instructions as to what Peter was to be taught and how it was to be documented. Howe was extremely dedicated, cropping her hair and even at one point painting the lower half of her face in thick white make-up and applying black lipstick so that Peter might more clearly see the shapes her lips made.

  Dolphins are promiscuous by nature, and as the experiment went on, Peter began to show a sexual interest in Howe; by week five she records in her notes that “Peter begins having erections and has them frequently when I play with him.” When she rebuffed his advances, Peter would become aggressive, using his flippers and nose to bruise her shins. When his conjugal visits with the other dolphins at the facility became so frequent as to be disruptive, Howe decided to relieve him manually so that he might focus on his lessons. Lilly, whose judgment may have been clouded by his incipient interest in LSD, noted, “I feel that we are in the midst of a new becoming; moving into a previous unknown . . .”

  Gregory Bateson, the distinguished linguist who served as the Institute’s director, and who was leading the research on dolphin to dolphin communication, was less than impressed. In his view, Peter was simply mimicking Margaret’s speech in order to get fish, with little or no comprehension of what was being said. Carl Sagan, too, was skeptical that any progress was being made and suggested that Lilly switch to experiments which could verify whether dolphins could convey information to one another rather than trying to teach them English.

  Eager for results, Lilly became increasingly desperate, and in the summer of 1966 he took the drastic step of injecting two of the dolphins with LSD to see whether that might improve their language skills. Thankfully, the drug appeared to have no effect, not even when Lilly took a pneumatic drill to the rock at the side of the pool to mess with the dolphins’ super-sensitive hearing. This new twist was too much for Bateson, who left the facility. Lilly’s NASA funding was withdrawn, the Institute closed down, and SETI was forever blemished by association.

  JUMPING THROUGH HOOPS

  The lesson of the Lilly experiments—apart from “don’t give dolphins handjobs or LSD”—is a basic one. When it comes to close encounters with other species, it’s all too easy to try to impose our own ideas of how they should think, feel, and behave. To us, a dolphin’s lack of facial expressions, for example, might seem to indicate a lack of empathy; to a dolphin, our reluctance to stand on our heads when we greet them might appear equally inscrutable. On one count, however, Lilly does appear to have been correct: Dolphins have all the trappings of high intelligence.

  For example, we know that not only are bottlenose dolphins capable of understanding signs, but the order in which they are shown those signs—in other words, the syntax—makes a difference. In the 1980s, Louis Herman and his team at the Dolphin Institute in Hawaii successfully taught captive dolphins “words” in the form of arm signals. Bottlenose dolphins were shown to be capable of understanding “sentences” of up to five “words,” easily distinguishing between “take the ball to the hoop” and “take the hoop to the ball.”

  As demonstrated by the work of Diana Reiss, bottlenose dolphins are capable of complex manipulation, creating underwater bubble-rings with their blowholes that they can then swim through; in one case a captive dolphin formed a stream of bubbles with its blowhole that it then shaped into a ring using its tail.24 They also show self-organized learning, interacting with a large poolside keyboard where they could request objects by pressing the relevant key, and exhibit self-awareness when presented with an underwater mirror, orienting their bodies to examine temporary marks made by Reiss and her fellow researchers.

  What was less clear, up until the late 1990s, was whether or not bottlenose dolphins’ wide range of vocalizations constituted any form of language. In the broadest of strokes, they make two main kinds of sounds with their blowholes: whistles and clicks. The clicks, generally speaking, are their equivalent of sonar, helping them to locate their prey and generally find their way around when visibility is poor.25 It’s the whistles—many of which can be outside the range of human hearing—where all the linguistic action appeared to be, but no one had any concrete proof.

  That was until the collaboration between SETI’s Laurance Doyle and UC Davis’s dolphin researchers Brenda McCowan and Sean Hanser. As an astronomer at SETI, Doyle had an interest in whether the mathematical techniques that were being used to pick out planets in the Kepler data could be used to decode alien messages. In particular, he was interested in what information theory, as devised by Claude Shannon, could tell you about an alien radio source. As you may remember from Chapter Five, Shannon showed that the maximum amount of information, H, contained in a message of i letters, each with probability pi, was simply:

  H = –∑ pi log2 pi

  In other words—and this is the only part you need to grasp—we can work out how much information could possibly be in a message just by knowing the letters the message is made up of, and how often they occur.

  If that sounds incredible, it should: To my mind, it’s a leap on the level of Einstein’s Theory of Relativity or Schrödinger’s Wave Equation. Let’s think back for a moment to the example we looked at in Chapter Five, where our apple-scrumping message is made up of two “letters,” “flashlight on” and “flashlight off.” Crucially, we calculated that the information contained in the message was 1 bit, even though we had no idea what that information was, i.e., whether you were coming scrumping or not.

  The team’s idea was simple. They had no SETI signal to look at, so why not look at the messages of dolphins using information theory to see how much information they were capable of carrying? And if they were capable of carrying a lot of information, was it anything like as much as human language?

  NAME, RANK, AND NUMBER

  Doyle knew McCowan through the Planetary Society, a US non-profit which maintains an avid interest in SETI. One of the problems with understanding dolphins is their prodigious acoustic ability: Water is much better at transmitting sound than air, and a dolphin can hear higher frequencies than a bat.26 To make matters even more complicated, dolphin whistles can be very short, lasting as little as a few tenths of a second, and change rapidly in pitch. As part of her PhD, McCowan had developed software that could not only accurately sample the individual whistles but could then categorize them by type. Now that there was a way of identifying the different “letters” in a dolphin message, could they be analyzed using information theory?

  One of McCowan and Hanser’s recent papers contained a table listing a number of different dolphin whistles and how often they had occurred, and to get things rolling the team decided to analyze them with a Zipf plot. For each of the forty-odd whistles in the table, they noted rank and frequency. Next they plotted them on a graph. Zipf’s Law, you’ll remember, can be written as:

  rank = constant/(frequency)

  or

  rank = cons
tant × (frequency)–1

  For reasons that shall soon become clear, let’s rewrite this as

  rank = (frequency)–1 × constant

  This kind of relationship is known in the trade as a power law, meaning simply that one variable (rank) is related to another (frequency) by a power index, in this case -1.

  That’s no fun to plot on a graph, so a standard trick in physics is to take the logarithm of both sides of the equation:

  Log (rank) = log { (frequency)–1 × constant} log (rank) = – log (frequency) + log (constant)

  Which is of that famous form, y = mx + c, the equation of a straight line. Plotting the logarithm of each whistle’s rank against the logarithm of its frequency, the team was astonished. It came out as a dead straight line with a backwards-sloping gradient of –1.00, like this:

  THE PLOT THICKENS

  Unbelievably, dolphin whistles meet one of the basic requirements of symbolic language: Like Ulysses, the Iliad, and Plains Cree, they obey Zipf’s Law. A result like that needs checking, so next the team looked at the Zipf plots of baby dolphins. Sure enough, the line ran with a flatter slope, indicating that a much wider variety of whistles was being used; in other words, the baby dolphins were babbling. Next, they plotted the whistles of dolphins between two and eight months old. This time, the Zipf line steepened to a –1.05 slope, showing that the toddler dolphins were repeating themselves. Finally, at between nine and twelve months, teenage dolphins plotted at –1.00, just like the adults. They had finally started making sense, no doubt earnestly informing one another that their parent dolphins were total losers.

  Next, the team decided to check how the calls of other species compared with bottlenose dolphins. Squirrel monkeys are highly social New World monkeys from Central and South America which make alarm calls to warn one another of predators, and Belding’s ground squirrels from the western United States chirp to warn one another of danger. Adults of both species were recorded, and their vocalizations analyzed on a Zipf plot. The squirrel monkeys plotted at a less impressive but still noteworthy –0.75, and the ground squirrels at a measly –0.30. Whatever the Zipf plot was measuring, adult bottlenose dolphins and humans had a lot of it, squirrel monkeys had a bit, and ground squirrels virtually none at all.

  SO COME ON, DO DOLPHINS HAVE LANGUAGE?

  As we’ve already mentioned, a Zipf plot is a necessary but not sufficient requirement for symbolic language, and it’s fairly easy to see why. Ulysses follows a Zipf plot, but it would be a brave literary theorist indeed that claimed that much of the complexity in Joyce’s masterpiece was captured by the knowledge of how often each word occurs.

  Clearly there is a lot more going on, and luckily Shannon has a fair bit to say about that, too. To write Ulysses, you not only need the right words, but you also need to put them in the right order. Looking at the text at the start of this chapter, for example, we can see that “of the” occurs five times; “of our” occurs three times. These two-word chunks are known as “digrams.” Clearly we need a lot more text to make a decent job of it, but you can see that we could do the same job for digrams as we did for individual words, working out the probability of each for Ulysses as a whole and plugging them into one of Claude Shannon’s—admittedly fierce—equations.

  If you do that, the number that you end up with is called a message’s “second order Shannon entropy.” In fact you can do the same sort of job with trigrams, quadrigrams, quintigrams, and so on, producing figures for the third, fourth, fifth, and higher Shannon entropies that capture more and more of the complexity of the message. Human language, for example, has a Shannon entropy of around eight, meaning that, on average, rules of syntax connect eight words at a time.27 And the dolphin whistles? They plotted at four, round about the same number of symbols that Louis Herman showed dolphins were capable of handling when they learned sign language.28, 29

  TALK TO THE ANIMALS

  So that’s it, right? Dolphins have language, but it’s not nearly as complex as ours. The answer, as you might suspect, is “not quite.” Firstly, we are measuring Shannon entropy, which tells us the potential for a message to contain information. Like a Zipf plot, a fourth-order Shannon entropy is a necessary-but-not-sufficient condition for complex language. Dolphins’ whistles might obey complicated laws of syntax without actually meaning anything, although given their highly social behavior that seems unlikely. Just as with the Ancient Egyptian hieroglyphs, until we manage to decipher bottlenose dolphin whistles we can’t be sure.

  And, second, a fourth-order Shannon entropy might not be the maximum entropy of dolphin whistles. As we know, to be able to calculate higher order entropies you need more and more text.30 Doyle, McCowan, and Hanser only had around 10,000 whistles to work with, so fourth-order entropy was the highest they could expect to find. You know you’ve reached maximum entropy when each n-gram—quadrigram, in this case—appears roughly the same number of times. Some dolphin quadrigrams, however, appeared a lot more often than others, which implies that if Doyle and his team had more whistle data they might have found higher orders of entropy. For all we know, dolphin whistles might have maxed out at eighth- or ninth-order, implying that—shock, horror—not only can they talk, but they are smarter than us.

  HOMO SAPIENS PHONE HOME

  In my humble opinion, the work of Doyle, McCowan, and Hanser has huge implications for any alien transmission we receive. Assuming we find a way to extract a message from it, one of the first things we are going to want to do is analyze it using information theory. That way, without knowing anything about what’s in the message—“We buy gold,” for example—we can figure out its potential to carry meaning. A fifteenth-order Shannon entropy would have our cryptographers quaking in their Birkenstocks, but at least we would know what we were dealing with.

  For us to be able to determine a fifteenth-order Shannon entropy, of course, we would need a whole lot of message. We’d also want it to be as diverse as possible, incorporating every conceivable kind of medium and exploring every nook and cranny of their culture. We’d want to hear alien music, flick through alien holiday snaps, and watch alien movies. After all, without a Rosetta Stone, decoding it is going to be prodigiously difficult. All we can hope for is some random cultural overlap—that we both spend long years raising children, for example, or share a love of selfies—so that we can find the equivalent of a cartouche: a small section of code for which the meaning is clear.

  In short, we want the aliens to send us their internet.

  METI-PHYSICS

  So, finally, if we are to send a message, what should it be? In fact, should we even signal at all? Many eminent scientists, Stephen Hawking among them, believe we shouldn’t. Contact with a more technological civilization didn’t turn out too well for the Plains Indians, he points out. To a certain extent, he’s right; we don’t know what’s out there. Maybe an advanced civilization will travel here by wormhole and suck up the Pacific into a giant water tanker.

  As Seth Shostak, the director of SETI, recently pointed out in a New York Times article, that’s a concern we never used to have. As we know, the Mir Message in 1962 set the ball rolling, with a brief Morse code message to Venus. Next came the so-called Golden Plaque message on Pioneers 10 and 11, launched in 1972 and 1973. The following year we sent the most powerful message we have ever transmitted, the Arecibo Message, which was fired at a cluster of some 300,000 stars known as Messier 13. Depicting a stick man, a twist of DNA, and a map of the solar system, the cruddy pixilation of the Arecibo Message makes the 1970s video game Pong look sophisticated.

  That was followed by the Voyager 1 and 2 probes in 1977 and the famous Golden Record. On it, as we’ve heard, Ann Druyan and Carl Sagan curated just about every kind of information they could get their hands on: speech, whale song, classical music, rock and roll, as well as images of the Taj Mahal and the underside of a crocodile. What they excluded, of course, was anything to do with war, politics, or religion. After all, we didn’t want to disappoint th
e aliens with our bad behavior.

  Since the Golden Record, the Crimea has become the main focus for Messaging to Extraterrestrial Intelligence, or METI, as it is now known thanks to the Russian astronomer Aleksander Zaitsev. In 1999 and 2003 he supervised the sending of two messages known as the Cosmic Calls to nine nearby stars. And, not to be outdone, in 2008 NASA beamed the Beatles’ song “Across the Universe” at Polaris, also known as the North Star.

  Rather than continuing to send such “greeting cards,” Shostak has a radical proposal, and one I heartily agree with: We should start sending “Big Data.” As Laurance Doyle’s work shows, we don’t need to overthink this. The first thing the aliens will want to do is work out whether there might be any information in our message, and to do that they will need a lot of data. There are other good reasons, too. In sending them everything we’ve got, we are being as honest as we can be about who we are, warts and all. Let’s not pretend we are sages, or saints. Let’s be human.

  We don’t have to worry about overloading them. To a technological civilization more advanced than our own, the world’s several hundred exabytes of stored information is going to seem like a flash drive.31 At the same time, while their technology might be more advanced, let’s not assume that they themselves are necessarily smarter than us. It’s our ability to manipulate information that has enabled our ascendance, not our individual smarts.