What Technology Wants (48 page)

Still, charismatic intelligence is relatively rare. But smartness is a competitive advantage everywhere. We see the widespread recurrence and reinvention of intelligence because the living universe is a place where learning makes a difference. Up and down the six kingdoms of life, minds have evolved many times. So many times, in fact, that minds seem inevitable. Yet as inordinately fond as nature is of minds, the technium is even more so. The technium is rigged to birth minds. All the inventions we have constructed to assist our own minds—our many storage devices, signal processing, flows of information, and distributed communication networks—all these are also essential ingredients for producing new minds. And so new minds spawn in the technium in inordinate degrees. Technology wants mindfulness.

This yearning for increasing sentience reveals itself in three different ways in the technium:

The technium is primed to hijack matter and rearrange its atoms to infiltrate it with sentience. There seems to be no place a mind can't be born or inserted. These mind children will be small, dim, and dumb at first, but tiny minds keep getting better and more abundant. In 2009 there were 1 billion electronic “brains” etched into silicon. Many of these tiny minds contain a billion transistors each, which the global semiconductor industry is manufacturing at the speed of 30 billion per second! The smallest silicon brain has a minimum of 100,000 transistors, about as many neurons as the brain of the rock ant. They, too, can do surprising feats. Tiny synthetic minds no bigger than an ant's know where on Earth they are and how to get back to your home (GPS); they remember the names of your friends and translate foreign languages. These dim minds are finding their way into everything: shoes, door-bells, books, lamps, pets, beds, clothes, cars, light switches, kitchen appliances, and toys. If the technium continues to prevail, some level of sentience will find its way into everything it creates. The smallest bolt or plastic knob will contain as many decision-making circuits as a worm, elevating it from the inert to the animate. Unlike the billions of minds in the wild, the best of these technological minds (in aggregate) are getting smarter by the year.

We are blind to this massive eruption of minds into the technium because humans have a chauvinistic bias against any kind of intelligence that does not precisely mirror our own. Unless an artificial mind behaves exactly like a human one, we don't count it as intelligent. Sometimes we dismiss it by calling it “machine learning.” So while we weren't watching, billions of tiny, insectlike artificial minds spawned deep into the technium, doing invisible, low-profile chores like reliably detecting credit-card fraud or filtering e-mail spam or reading text from documents. These proliferating microminds run speech recognition on the phone, assist in crucial medical diagnosis, aid stock-market analysis, power fuzzy-logic appliances, and guide automatic gearshifts and brakes in cars. A few experimental minds can even drive a car autonomously for a hundred miles.

The future of the technium at first seems to point to bigger brains. But a bigger computer is not necessarily smarter, more sentient. And even when intelligence is demonstrably greater in biological minds, it is only weakly correlated to how many brain cells are present. In nature, animal computers come in all sizes. An ant brain is a 100th-of-a-gram speck; the 8-kilogram brain of a sperm whale is 100,000 times bigger. But it is not clear that a whale is 100,000 times smarter than an ant or that humans are only three times as smart as chimpanzees, as the specifications of pure numbers of cells might suggest. Our large human brain, with its endless ideas, is only one-sixth the size of a sperm whale brain. It is even slightly smaller than the average Neanderthal brain. On the other hand, recently discovered minihumans on Flores Island had brains one-third the size of ours and may have been no dumber. The correlation between the absolute scale of the brain and smartness is not significant.

The architecture of our own brain suggests the future of artificial sentience may reside in a different kind of big. Until recently, conventional wisdom held that specialized big-brain supercomputers would first host artificial intelligences, and then perhaps we'd get mini ones at home or add them to the heads of our personal robots. They would be bounded entities. We would know where our thoughts ended and theirs began.

However, the snowballing success of search engines such as Google this past decade suggests the coming AI will most likely not be confined in a stand-alone supercomputer but will be birthed in the superorganism of a billion CPUs known as the web. It will run on the global megacomputer that encompasses the internet, all its services, all peripheral chips and affiliated devices from scanners to satellites, and the billions of human minds entangled in this global network. Any device that touches this web AI will share—and contribute to—its intelligence.

This gargantuan machine already exists in a primitive form today. Consider the virtual supercomputer of all the world's computers online. There are one billion online PCs, which is about as many transistors as are in an Intel chip in one computer. All the transistors in all the computers connected together add up to about 100 quadrillion (10
¹⁷
) transistors. In many ways, this global virtual network acts like a very large computer that operates at approximately the clock speed of an early PC.

This supercomputer processes three million e-mails each second, which essentially means network e-mail runs at 3 megahertz. Instant messaging runs at 162 kilohertz, SMS at 30 kilohertz. In any one second, 10 terabits of information can be coursing through its backbone, and each year it generates nearly 20 exabytes of data.

This planetary computer embraces more than just laptops. Today it contains approximately 2.7 billion cell phones, 1.3 billion land phones, 27 million data servers, and 80 million wireless PDAs. Each device is a differently shaped screen that peers into the global computer. It takes a billion windows to glimpse what it is thinking.

The web holds about a trillion pages. The human brain holds about a hundred billion neurons. Each biological neuron sprouts synaptic links to thousands of other neurons, while each web page on average links to 60 other pages. That adds up to a trillion “synapses” between the static pages on the web. The human brain has about 100 times that number of links—but brains are not doubling in size every few years. The global machine is.

And who is writing the software that makes this contraption useful and productive? We are, each of us, every day. When we post and then tag pictures on the community photo album Flickr, we are teaching the machine to give names to images. The thickening links between caption and picture form a neural net that can learn. Think of the 100 billion times
per day
humans click on one web page or another as a way of teaching the web what we think is important. Each time we forge a link between words, we teach it an idea. We think we are merely wasting time when we surf mindlessly or blog an item, but each time we click a link we strengthen a node somewhere in the supercomputer's mind, thereby programming the machine by using it.

Whatever the nature of this large-scale sentience, it won't even be recognized as intelligence at first. Its very ubiquity will hide it. We'll use its growing smartness for all kinds of humdrum chores—data mining, memory archive, simulations, forecasting, pattern matching—but because the smartness lives on thin bits of code spread across the globe in windowless boring warehouses, and it lacks a unified body, it will be faceless. You can reach this distributed intelligence in a million ways, through any digital screen anywhere on Earth, so it will be hard to say where it is. And because this synthetic intelligence is a combination of human intelligence (all past human learning, all current humans online) and digital memory, it will be difficult to pinpoint just what it is. Is it our memory or a consensual agreement? Are we searching it, or is it searching us?

Someday we might meet other intelligences in the galaxies. But long before then we will manufacture millions of new kinds of minds on our own world. This is the third vector of evolution's long-term trajectory toward increased sentience. First, insinuate intelligence into all matter. Second, bring all those embedded minds together. Third, increase the diversity of minds. There may be as many species of intelligence possible as there are species of beetles, which is saying a lot.

There are a million and one reasons to build a million and one different types of artificial intelligences. Specialized intelligences will perform specialized tasks; other AIs will be general-purpose intelligences that accomplish familiar tasks differently from how we do. Why? Because difference makes progress. The one kind of mind I doubt we'll make many of is an artificial mind just like a human. The only way to reconstruct a viable human species of mind is to use tissue and cells—and why bother when making human babies is so easy?

Some problems will require multiple
kinds
of minds to crack, and our job will be to discover new methods of thinking and to set this diversity of intelligences loose in the universe. Planetary-scale problems will require some kind of planetary-scale mind; complex networks made of trillions of active nodes will require network intelligences; routine mechanical operations will need nonhuman precision in calculations. Since our own brains are such poor thinkers in terms of probability, we'd really benefit by discovering an intelligence at ease with statistics.

We'll need all varieties of thinking tools. An off-the-grid stand-alone AI will be handicapped compared with a hive-mind supercomputer. It can't learn as fast, as broadly, or as smartly as one plugged into six billion human minds, several quintillion online transistors, hundreds of exabytes of real-life data, and the self-correcting feedback loops of the entire civilization. But a consumer may still choose to pay the penalty of lesser smarts in order to have the mobility of an isolated AI in distant places, or for privacy reasons.

Currently we are prejudiced against machines, because all the machines we have met so far have been uninteresting. As they gain in sentience, that won't be true. But we won't find all types of artificial minds equally attractive. Just as we find some natural creatures more charismatic than others, some minds will be charismatic (attractive to our way of thinking) and some won't. In fact, we might be repulsed by the alien nature of many of the most powerful types of intelligences. For instance, the ability to remember
everything
can be scary.

What technology wants is increasing sentience. This does not mean evolution will move us only toward one universal supermind. Rather, over the course of time the technium tends to self-organize into as many varieties of mind as is possible.

The primary thrust of exotropy is to uncover the full diversity of intelligences. Each type of thinking, no matter how large it is scaled up, can only understand in a limited way. The universe is so huge, so vast in its available mysteries, that it will require every possible type of mind to comprehend it. The technium's job is to invent a million, or a billion, varieties of comprehension.

This is not as mystical as it sounds. Minds are highly evolved ways of structuring the bits of information that form reality. That is what we mean when we say a mind understands; it generates order. As exotropy pushes through history, self-organizing matter and energy into greater complexity and possibilities, minds are the fastest, most efficient, most exploratory technology so far for creating order. By now our planet owns the dim minds of plants, the multiple manifestations of a common animal mind, and the restless self-consciousness of human minds. Just a second ago, cosmically speaking, human minds began to invent a second generation of sentience. They installed their inventiveness in the most powerful force in the world—technology—and are trying to clone their own tricks. Most of these newly invented minds are no more intelligent than plants, a few are as smart as insects, and a couple hint at greater thoughts to come. All the while, the technium assembles brain-like networks at scales way beyond individual humans.

The trajectory of the technium is pointed toward a million more minds inhabiting the least bits of matter, in a million new varieties of thinking, subsumed with our own multiple minds into a planetary thought—on the way to comprehending itself.

It took Sapiens several million years to evolve from an apelike ancestor. During that transition to humanity, our DNA changed by a few million bits. So the natural rate of biological evolution in humans, in terms of information accumulation, is about one bit per year. Now, after almost four billion years of bit-by-bit biological evolution, we have unleashed a new type of evolution, one that creates rivers of mutations using language, writing, printing, and tools—what we call technology. Compared to the one bit per year we made as apes, we are adding 400 exabytes of new information to the technium each year, so the rate of our technological evolution is a billion billion times as fast as the evolution of DNA. As humans it takes us less than a second to process the same amount of information that our DNA took a billion years to process.

We are accumulating information so rapidly that it is the fastest increasing quantity on this planet. The amount of mail sent through the U.S. postal system has been doubling every 20 years for 80 years. The number of photographic images (a very dense information platform) has risen exponentially since the medium was invented in the 1850s. The total number of telephone-call minutes each day likewise has followed an exponential curve for over 100 years. There's no stream of information that is lessening.