The Computers of Star Trek (16 page)

Which returns us to the question we asked early in this chapter. Would it be possible for Data to play the violin? Or bend, twist, and turn while dancing? How can a machine operate with equal or better skill than a man?

Simple. Combine microhydraulics in a body that doesn't wear out or become tired with the information storage capacity of a thousand Libraries of Congress. Mix the knowledge and motor speeds together in a control unit that processes 60 billion operations per second, and you have a dancing, fencing, walking, talking, singing, acting, typing android. It's impossible using current science, but researchers today are making tremendous strides in robotics, neural nets, and nanotechnology. Couple these strides with three more centuries of advances, and you have a humanoid robot capable of performing every physical act displayed by Data on
Star Trek
. However, Data's more than a robot. He's a sentient being. Data thinks.

Perhaps the most famous test of whether a computer is artificially intelligent was proposed by British computer scientist Alan Turing. In its simplest form, Turing stated that a computer is intelligent if it deceives a human into believing it's human. Turing's approach is a test of top-down AI.

Data has passed Turing's test many times, with inhabitants of other planets as well as his own crew. Since Data is obviously artificially intelligent with a positronic neural net, is he “top down” or “bottom up”? Or a combination of both?

Data's mind appears to be a “bottom-up” Al/neural network. He's not loaded with facts about everything in the universe (while his memory is large enough to hold the information in a thousand Libraries of Congress, it's not as huge as the one in the ship's computer). Data possesses a great deal of information. In “Birthright,”
TNG
, he mentions that he's analyzed over 4,000 different religious and philosophical systems, but it is implied that this knowledge has been acquired by high-speed reading. He's shown constantly learning, growing, and evolving over a period of time. As an artificial being, Data's not shy and asks for advice and opinions, often from a long-suffering Geordi LaForge. Data has a cat named Spot and tries constantly to understand it. He even dreams and tries to learn from these dreams. Only a bottom-up creation would try to understand both romance and humor.

Data operates in the holosuite, an artificial environment, as well as the real world environment, without any problem whatsoever. (“Elementary, Dear Data,”
TNG, Generations
, and many more episodes) Again, coping with constant novelty implies bottom-up AI neural networking.

Still, Data (along with Lore and Mrs. Soong) seems to possess vast amounts of practical and esoteric information. His mind is full of facts. Data's a great fan of Sherlock Holmes, which implies heuristic thinking and a top-down Al approach. In many
Trek
adventures, Data uses the if-then type of thinking associated with logic trees to arrive at a conclusion. More telling, to function as a highly rated Federation officer, Data must make snap decisions and do snap calculations. His mind obviously contains a huge amount of information that helps him reach these conclusions. In
Insurrection,
Data comments that if one of his legs grows, he won't be able to walk. This implies that his walking algorithms are top-down AI—they can't adapt and let Data limp. When confronted with new stimuli, Data doesn't over-react. He never
breaks down. If anything, he adapts more quickly than any other crewmember (“Cause and Effect,”
TNG
). His learning phase is incredibly short.

Data deals with concepts that are simple for ordinary human brains but extremely complex for artificial neural nets. What we do naturally is almost impossible to duplicate in a computer. Let's consider, for example, vision, a key issue for creating androids that interact with their environments or survive hostile situations.

The sensory abilities and neural architectures of animals tend to be based on what they need. One factor is reaction time—how long do I have to drop this chunk of meat and run before the dog sees me? Other factors are the sizes and complexities of the objects that the animal recognizes.

Human visual perception is based on much more complex input. The retina uses cells called rods to handle incoming light and cells called cones to handle incoming color. With approximately 100 million rods and cones, the retina processes images at the rate of something like 10 billion calculations per second.

After image preprocessing by the retina, the cerebral cortex takes over the image processing. Vision centers for this purpose occupy more than half of the cortex. At this point, our brain hasn't even started to determine what it is we're looking at—a ball, a field of flowers, or a crowd of people—much less fit all the objects into a moving scene, analyze who and what we know, or how we want to react. All our brain has right now is raw image data. Think of it as the collection of individual bits accumulated by a digital camera before you get to see the image, the photo, you just took.

But our visual skills are far more elaborate than the world's finest cameras. We instantly recognize and respond to textures, lights, and shades. Our brains create an imprint that is complete with texture, object boundaries, and compensation for fog, brightness, and shadow. Simply by looking at the leaves on a
plant, we deduce that they are prickly, smooth, sticky, dangerous, or safe to touch. In dim or flickering light, we still recognize our friends or our enemies. Just the outline of someone's profile in heavy fog may suffice for visual recognition. We even interpret and process oddities such as optical illusions. If a satanic head appears to us in a cloud, we might conclude that—well, perhaps we need a bit more sleep.

It's unlikely that Data can process optical illusions. He might recognize Riker or Geordi from the back, he might be able to handle complex three-dimensional image processing, but he doesn't have the requisite innate human ability to process illusions or shifts in shadows and lights. His positronic neural net can learn using back propagation; in other words, by making mistakes. He might think that the satanic head in a cloud is real. For a better example, suppose the Borg, who are digitally driven creatures themselves, send Data's positronic brain some wireless visual stimulation that makes Data see all crewmembers as Borg. How would he interpret this visual information? In a dangerous situation, he might accidentally shoot Picard, thinking he's killing a Borg. Only by making a mistake—in this case a serious one—would Data learn that the Borg have impressed optical illusions into his visual system.

While Data's learning the hard way, we fear that he'd probably die. After all, he often functions under life-threatening circumstances. Is the shift of light in the corridor an alien attack or something else?

Data shows his ability to learn from his mistakes, a prime indicator of “bottom-up” artificial intelligence, in the adventure “Peak Performance.” He finds himself matched in computer strategy games against an arrogant Zakdom tactician, Sima Koirami. Data loses and believes his programs must be faulty. He warns Captain Picard that his judgment might be inaccurate. He even resorts to studying his own schematics to locate the problem.
It's inconceivable to him that he might lose a computer strategy game to a living being. Data's initial reactions and efforts are “top-down,” as he tries to use logic-trees and “if-then” reasoning to discover why he lost. But none of his attempts bring success. It's not until he studies the problem and learns from his previous mistakes—typical “bottom-up” AI thinking—that he finally comes up with a strategy to defeat Koirami.

We can conclude that Data has “bottom-up” AI, but with more than a touch of “top-down” AI tossed in for good measure.

Androids like Data aren't just a few years away from now. We don't expect to see them strolling down the street anytime soon. Still, with the rapid developments taking place in neural nets and AI, it seems likely that beings like Data might exist well before the
Enterprise-D
ever leaves on its first mission.

In his attempt to become more human, Data's most difficult task has been to feel and understand human emotions. He has tried to understand humor (“The Outrageous Okona,”
TNG
), romance (“In Theory,”
TNG
), and even parenthood (“The Offspring,”
TNG
). A submerged subroutine, triggered by a plasma shock, enables Data to dream. (“Birthright, Part 1”
TNG
) Originally unable to experience emotions, Data later obtained an emotion chip created by Dr. Soong and gained the ability to turn his emotions on and off (“Descent, Part 2,”
TNG, Generations, First Contact
).

Apparently, the chip enables Data to feel real emotions, not computational facsimiles. When coding emotion into virtual humans, sophisticated programs may generate facial expressions and body postures that simulate emotional reactions by real people. The key word is simulate. The emotion chip does more than that for Data. It allows him to experience true emotion.

In
First Contact
, the Borg hive-queen activates Data's emotion chip and grafts a patch of organic skin onto the endoskeletal
structure of Data's arm. She then blows on the skin, and Data has some kind of orgasmic upswing. A few moments later, while he is cradling the skin protectively, the Borg queen challenges him to tear it off his body. He cannot bring himself to do it. Apparently, Data's normal skin (the skin “fabric” that the nanites penetrate in “Evolution,” [
TNG
]) doesn't have nerve endings that feed into his emotion chip. But his new organic skin does.

While it's clear that the movie is making a philosophical point—without an organic body, emotions aren't really possible—as a matter of simple wiring, the scene doesn't make much sense. If the internal connections between Data's normal skin and his emotion chip didn't exist before, it's hard to see how grafting a new patch of organic skin onto him could create them. Unless he has millions of unused, superfluous “nerve” endings lying just below his normal covering.

We think the movie is wrong: An organic body is not a prerequisite to emotions. Emotions are a product of evolution—a powerful adaptation that helps us learn and survive—and there's no reason why an evolved alife intelligence couldn't have them, too. Some estimates place emotional reactions in artificially intelligent computer systems by the year 2050. It all depends on how quickly the next computer revolution occurs—the one that lifts us from microprocessors to DNA, quantum, optical, holographic, and other forms of computer technology. Clearly, by Data's time, he won't require a special “emotion chip.” In three or four hundred years, androids will have built-in, automatic emotional responses. The real Data will have a wide range of emotions, unlike the
Trek
Data with his static qualities of the galaxy's finest boy scout.

There are many ways to relax on the
Enterprise.
People listen to music, read (one of Captain Picard's favorite activities), engage in dramatic performances, or even work up a sweat in sports. Yet without question, the holodeck is the most advanced form of entertainment ever invented.

It's on the holodeck that our favorite characters act out their most cherished dreams. The holodeck can provide an exotic background for a social gathering or settings and action for lifelike interactive novels, such as the ones where Captain Picard plays the hard-boiled detective Dixon Hill. Julian Bashir plays at being an absurdly dashing secret agent, Tom Paris reinvents himself as a character in a twentieth-century science-fiction epic, and we have no doubt that the holosuite programs in Quark's library range from the romantic to the extraterrestrially revolting. Holodecks and holosuites offer a chance for relaxation, entertainment, or adventure. Unfortunately, they're also extremely unlikely to exist, even three hundred years from now.

The holodeck (we'll stick to holodecks, but everything we say about them applies equally to holosuites) relies on perhaps the most complex computer program on the
Enterprise
, as well as on transporter and replicator technology. We're told explicitly that
the holodeck is a direct outgrowth of transporter technology (“Heroes and Demons,” VGR). Transporters, replicators, and holodecks all rely on the ability to assemble matter at the molecular level, either from a template that was disassembled elsewhere just moments before, a pattern stored in memory, or a set of general instructions. Furthermore, we're told that unlike the replicators, the holodecks create solid-seeming objects out of some sort of magnetic pseudo-particles rather than real matter. This technology clearly raises questions of plausibility, many of which are best answered by a physicist. We'll stick to computing issues.

According to the
Star Trek: The Next Generation Technical Manual,
¹
the
Enterprise
contains four main holodecks, located on Deck 11. Twenty smaller holographic units (probably similar to the holosuites in Quark's bar) are on Decks 12 and 33. The holodecks use holograms and replicator technology to create realistic and believable simulations.

The holodeck creates illusions in a variety of ways. The gridlike walls can generate images of immense distance, such as the ocean in
Generations
or the crowded vastness of a nineteenth-century London cityscape in “Elementary, My Dear Data” (TNG). Holograms are routinely projected onto the deck for scenery, creating everything from landscapes to ancient fortresses. Most of these background features, and the characters moving through them in non-active roles, have no need for physical form and are obviously mere projections. These effects are merely extensions of today's virtual reality programs.