The Future of Computers


(c) 1996 Robert A. Freitas Jr.

Research Scientist

Zyvex Corp.


Citation:  Robert A. Freitas Jr., “The Future of Computers,” Analog 116(March 1996):57-73.


(2 November 2002:  I began work on this collection of  speculative ideas in late 1994 as a newcomer to the field, having read only Drexler’s three books and being completely ignorant of other similar speculations published elsewhere – which would have been cited, had I known of them – that in some cases anticipated some of my thoughts.  The article was submitted on 15 March 1995, accepted for publication on 12 April 1995, and finally published in January 1996.)





This year's big supercomputing conference, "SuperComputing 95," to be held in San Diego, CA in December 1995, features what sponsors are excitedly calling the Teraflop Challenge.  This is a contest to see if supercomputer designers can prove that one of their latest machines can compute at the peak speed of one trillion Floating-point Operations per Second (FLOPS).  It's an even bet someone will win.


One teraflop.  That's One Human, folks.


Yes, I know you need software.  Also, the processing capacity of the brain has not been reliably determined.  But a fair estimate is that the 1.5 kilogram organ has 1010 neurons with 103 synapses firing an average 10 times per second, which is about 1014 bits/second.  Using 64-bit words like the largest supercomputers, that's about one teraflop.


Figure 1 charts a century of evolution in data processing power based on more than 100 mechanical and electronic computer systems.  I started with Moravec's numbers (Ref. 1) and added new information for machines built since 1987, various robot systems and NASA computers I'm familiar with, and a few mass-consumer products we have around the office.  I also included a few projected machines:  The 9-teraflop Cray/DARPA (Defense Advanced Research Projects Agency) machine, due in 1998;  the 1-teraflop NEC Super, due in 2000;  and the proposed 2-teraflop upgrade of Thinking Machines Corporation's Connection Machine-5, which the company claims could be assembled by 1996 using existing technology for a stiff $100 million fee.


Figure 2 shows the declining cost of teraflop capability in constant (inflation-adjusted) 1995 dollars.  According to these data, by 2001 the price of a Human-Equivalent Computer (HEC) should fall to one million dollars, well within the research budgets of most universities and research labs.  The 21st century will open with an explosion of intensive experimental and development work to create "man-in-a-box" software.  By 2011, teraflop systems should cost only $1000 and will begin to be incorporated into homes, cars, and high-end household appliances.  Astonishingly, Figure 2 suggests that by 2021, HECs will cost a mere $1, which means they'll start showing up in children's toys and magazine inserts.  Thus the chart tells us that just 25 years from today, human-equivalent artificial intelligence (AI) may become "too cheap to meter."






To keep the progression going, we need faster and faster components.  Intel's successor to the Pentium, the P6 chip, clocks in at 133 million ops/sec.  Texas Instruments' new superchip, the MVP, handles two billion digital signal processing operations a second (a 0.002 teraflop device).  Using conventional silicon technology, if DRAM (Dynamic Random Access) memories keep doubling their surface density every 2 years, as they have for decades, then by 2001 we'll have a gigabit storage chip, pushing silicon near its ultimate limit.  (At 0.1-micron, about the wavelength of an electron, silicon transistors break down as electrons start tunnelling through their switchgates.)


Then what?  Mark Johnson at Bellcore in Red Bank, New Jersey, proposes building computers with "bipolar spin switches" (Refs. 2,3).  Johnson's switch, which uses the direction of a magnetic field to control an electric current, is a sandwich of two magnetic alloys and a film of gold.  The first magnetic film polarizes the spin axes of electrons carrying electric current in the adjacent gold layer.  A population of spin-polarized electrons builds up in the gold, causing a "spin bottleneck" effect.  The second magnetic layer opens the switch (allowing electrons to flow) when its magnetism parallels the first layer's, and closes the switch when the magnetism is reversed.  Unlike semiconductor transistors, the performance of all-metal spin switches should continue to improve well below the 0.1-micron silicon limit.


A better-studied alternative is the Josephson junction, a niobium/metal-oxide superconducting electronic switching device first investigated by IBM in the 1960s.  A Josephson logic gate operates much faster but differently than a semiconductor gate – it switches like a latch and does not return to its initial state after the input signal is turned off.  By comparison, a semiconductor gate changes state only while the input signal is applied.


The first stable, high-quality Josephson junction was fabricated in 1983.  By 1989, Japanese researchers at Electrotechnical Laboratory in Tsukuba had constructed a prototype 4-bit superconducting computer named ETL-JC1.  This was the first machine to execute computer programs installed in a Josephson ROM (Read-Only Memory) chip with a Josephson central processing unit and RAM (Random Access Memory) chips.  By 1990, the Japanese group advanced to an 8-bit microprocessor with 6300 gates operating at 770 MHz clock speeds. (Ref. 4)


Konstantin Likharev of the State University of New York at Stony Brook is pushing the technology even further using a new kind of gate logic called Rapid Single Flux Quantum (RSFQ) logic, that promises to overcome drawbacks of earlier strategies.  Already, simple RSFQ circuits made of the low-temperature superconductor niobium are running at speeds of 50 billion cycles per second, 1000 times faster than Intel's Pentium and 100 times faster than the quickest silicon devices.  "Improving the technology," says Likharev, "offers hope to achieve speeds of around 300 billion operations a second.  Energy consumption is a million times less than typical silicon memory cells."


Optical computers are another option.  Alan Huang and colleagues constructed the world's first digital optical processor in 1990 using Symmetric Self-Electrooptic Devices (S-SEEDs), computing elements first invented at Bell Labs in 1987.  S-SEEDs are optical switches made of gallium arsenide built by molecular beam epitaxy, having a potential clock speed of 1 GHz and picojoule switching energies.  The devices are 5000 nm (nanometers) square and use two mirrors with controllable reflectivity to infrared light.  The demonstration processor, which operates at only 1 MHz, contains 32 S-SEEDs and two laser diodes in each of four arrays connected by lenses and masks serving as logic gates.


In 1993, University of Colorado researchers Vincent Heuring and Harry Jordan took the next step, unveiling the world's first general-purpose optical computer able to store and manipulate its own programming instructions internally.  The prototype has the power of a small personal computer (3 x 106 bits/sec).  But the inventors claim that using technologies already demonstrated in the laboratory they could build a palm-sized version 400 times smaller than the current prototype, and boost power into the 109 bit/sec range (Ref. 5).


Many strange ideas lurk on the horizon.  Back in the early 1960s, Rolf Landauer of IBM suggested that the only logic operations that necessarily require the dissipation of energy are those which are thermodynamically "irreversible."  That's how conventional computers work – logic circuits and other components are set up to run in one direction only.  Landauer's observation has led to several proposals for quantum computers employing (1) reversible, dissipationless logic devices, (2) computations carried out using reversible logic alone, and (3) bits that are registered by true quantum-mechanical quanta such as spins.


Testing the first of these proposals, in 1992 Xerox Corp. physicist Ralph Merkle and several collaborators sought to create more energy-efficient devices by rearranging a set of CMOS transistors into reversible switches (Ref. 3).  They interspersed inductors among the switches in order to harvest electrical energy that would have been lost as heat, and fed it back into the power supply.  The reversible circuits are 7.7 times more efficient than conventional ones, albeit much slower.  "My rash prediction," says Merkle, "is that reversible logic will dominate in the 21st century."


The latest in reversible devices is Seth Lloyd's proposal for a "quantum supercomputer" (Ref. 6).  The device, he explains, would be made of "arrays of weakly coupled quantum systems.  Computation is effected by subjecting the array to a sequence of electromagnetic pulses that induce transitions between locally defined quantum states."  For example, in one dimension the computer might consist of localized electronic states in a polymer;  in two dimensions, quantum dots in a semiconductor;  in three dimensions, nuclear spins in a crystal lattice.  Lloyd achieves quantum-mechanically coherent computation by "preparing the array in a superposition of program states and then operating the array as a digital computer."  The proposed device would not just be a universal digital computer, but also a general-purpose quantum-mechanical micromanipulator.






As computers progress toward human intelligence, robots are taking more human form.  "Cog," the latest project at Rodney Brooks' Artificial Intelligence Lab at MIT, may be the most ambitious humanoid robot ever attempted.  Researchers hope the machine will eventually approach, and in some areas perhaps surpass, the mental and physical prowess of a 6-month-old human infant (Ref. 7).  Using mechanical eyes, ears, arms, fingers, and a computer brain inspired by human neuroanatomy, the robot will tackle questions such as how eye-hand coordination develops in the brain and how infants learn to interact with others.  Complete realization of this goal could take a decade of work and millions of dollars.


Why is Cog so important?  In the 1980s, Brooks, the controversial "Bad Boy of Robotics," built more than two dozen small, fast insect-like robots challenging the conventional wisdom that mobile, autonomous machines must extract a symbolic representation of the world from their sensory data before they can plan a course of action.  Brooks' insect robots, in contrast, had a more direct coupling with the world, using simple sensors to continually modify and incite physical actions.  For example, when a leg of one of his insects hit an obstacle, a contact sensor quickly invoked a reflex action like "lift leg" or "back up."  Dozens of these stimulus-response reflexes were linked together in a rich web, giving Brooks' mini-robots surprisingly complex, robust – and quick – behavior.


During human evolution, Brooks says, the human brain must have developed thousands of solutions to everyday problems like seeing, hearing, and moving.  Higher intelligence may simply be a matter of adapting these solutions in a novel way.  "When you're building in the machinery that knows not to bump into things, how to get around them, et cetera, you're also building in a lot of the same machinery that does abstract reasoning," explains Lynn Stein, an MIT roboticist working with Brooks on the Cog project.  Why the sudden leap from robot bugs to robot men?  "I figure I have one more 10-year project left in me," laughs Brooks, who has labored in the AI vineyards for decades.  "I'd hate to go out as `he built the best artificial cat in the world'."


When might a Human-Equivalent Computer fit in a machine the mass of a human body?  The trendline on the chart in Figure 3, based on historical data from actual computers, implies a human-size teraflop system should arrive by the year 2007.  By 2025, HECs should weigh as little as a credit card, like "SELMA" in the TV series Time Trax.  By 2050, the curve reaches one microgram, a Human-Equivalent Computer the mass of a small paramecium.  As you'll see below, these last two projections are unduly pessimistic.






Molecular nanotechnology, a concept first articulated by Eric Drexler in the early 1980s, is the controlled manipulation of matter at the atomic and molecular levels to create new products with atom-by-atom precision.  (A nanometer is about 6 atoms wide.)  Drexler has since written three books on the subject (Refs. 8-10) with a fourth due out in 1996 [Ed. note -- this once-rumored book was never written], and there are many other excellent summaries of technical progress in this rapidly growing area (Refs. 1,11-15).  The field has its own journal, Nanotechnology, published quarterly since 1989.  Interested readers can keep abreast of the latest developments, books, and conferences by becoming a member of Drexler's organization, the Foresight Institute (Box 61058, Palo Alto, CA 94306, 415-324-2490).


Where are we today?  Engineers now manipulate matter at the atomic level, carving out 1-atom-deep "molecular corrals" and fabricating 1-10 nm scale "suspension bridges," "guitar strings," wires, magnets and bearings.  (Complex biomolecules like insulin and hemoglobin are compact natural nanomachines about 2-3 nm in size.)  The Atomic Force Microscope (AFM) and the Scanning Tunneling Microscope (STM) permit the precise positioning of individual atoms.  The newest machines cut the time required to etch 1-nm lines to seconds.  In 1992, researchers at Harvard University used an AFM to perform nanomachining operations on a molybdenum trioxide crystal (Ref. 16).  Using an applied load of 100 nanonewtons at the tip, they milled a triangular-shaped 50-nm part from the crystal, then slid the part 200 nm across the worksurface.  "Nanoparts" with features as narrow as 10 nm can be machined this way.


Smaller nanoparts can be constructed chemically.  Recently, Professor T. Ross Kelly and his colleagues at Boston College created a "paddlewheel" molecule (a spinning propeller-shaped wheel) with a built-in brake.  In solution, the wheel spins freely.  When chemists add mercury ions, the brake trips, stopping the rotation.  When they remove the mercury, the wheel resumes spinning (Ref. 17).


Nanoparts can also be hooked together chemically.  Last year two British chemists created a self-assembling molecule consisting of five interlocked rings (five nanoparts) averaging 75 atoms per ring, arranged in the shape of the Olympic logo.  This is the largest mechanically-interlocked molecule synthesized to date.


In a few years it will be possible to construct complex nanoparts consisting of a few thousand atoms, such as the planetary gear shown in Figure 4A, on a timescale of days.  Once we get more proficient at building nanoparts, we can put a few hundred of them together to build the first nanotools consisting of millions of atoms, much like the hypothetical manipulator arm (good for nanoassembly work) shown in Figure 4B. 


Nanoparts may also be assembled into nanomechanical logic devices.  We can use 1-nm sliding diamond rods to build gates (Figure 5A) and registers (Figure 5B) with 0.1-nanosecond switching speeds.*  These devices could be combined to make programmable logic arrays (Figure 5C) with >1 GHz clock speeds.  Drexler's benchmark nanocomputer with 100,000 logic rods, 10,000 registers, power supplies, etc. occupies a cube 400 nm on a side, weighs 10-7 micrograms, and computes at 1 billion ops/sec (0.001 teraflops) (Ref. 10).


[* Nanoelectronic systems should be much faster.  The ultimate:  A "nanooptical" computer limited by quantum theory to a clock speed of 1014 transitions/sec, a mid-infrared frequency at 0.4 eV, about 10% of carbon-carbon bond energy.  Attempting to clock faster than this would start tearing apart the switches.  Note that a "nanooptical" computer is made of human-equivalent gates!]


How soon might we be able to build Drexler's mechanical nanocomputer?  Figure 6 shows the evolution of data processing power per unit volume over the last 100 years.  Drexler's design represents a computing power density of 5 x 1029 bits/sec/cubic meter, assuming 32-bit words.  Figure 6 says this density should be reached by 2025, but we could see the first simple nanocomputer, performing just 1000 ops/sec in a 100-nm volume, as early as 2015.






I've collected the above projections into a single timeline (Figure 7).  It looks like we'll be nearing the physical limits of nanotechnology by mid-21st century, a time when nanorobots and Human-Equivalent Computers will be as ubiquitous as electricity, indoor plumbing, ballpoint pens and automobiles are today.


Lots of interesting applications have already been mentioned in print.  A few notables:  Video paint;  self-repairing mountain-climbing rope;  shape-changing cars, houses, furniture and clothing;  self-cleaning carpets and windows;  conscious control of pregnancy, sleep, and physical appearance (metamorphic cosmetics, programmable tattoos);  complete automation of personal hygiene including hair, skin, teeth, and bowels (use your imagination);  nanofood (synthohol, programmable flavors);  nanodrugs (Library pills, externally switchable analgesics and aphrodisiacs, rejuvenant "Venus" pills);  self-assembling critical masses of radionuclide-rich nanorobots (atomic stealthbombs);  and the ability to read, copy, rewrite, add, erase, transfer, or exchange thoughts and memories to the brain, with or without cooperation (lie detection, mental rehabilitation).


By the mid-21st century, nanotechnology will make possible many science fiction inventions once relegated to the distant future.  Very little lies beyond our grasp.  Figure 8 summarizes the characteristics of representative nanomachine systems as they are likely to exist by 2040, using parameters derived from Drexler's textbook (Ref. 10).  The following extrapolations strictly conform to all gross engineering specifications in the table.  As is common in exploratory engineering exercises, we attempt to make the case that certain future systems are physically possible, not to actually design those systems today.


The results – to me – are simply astounding.


TRACTOR BEAMS.  Consider a tractor beam in space, designed to hold, say, a 1000-ton object motionless against a 1-gee shearing force, 1 km from your ship.  Rather than some unspecified and mysterious beam, we use a physical cable with a 1-centimeter diamond core coated with a 10-micron layer of 0.1-1.0 micron nanorobots.  The diamond core is in short segments so the hawser stays flexible during storage and deployment.


A 1-km length of coiled cable weighing 300 kg is paid out from a beachball-sized pod fired at 1 km/sec towards the target.  Using windlass brakes the pod decelerates and impacts the target at low speed.  Pod nanorobots secure it to the object with diamond microanchors and molecular welding.  Assembler nanorobots in the coating weld the diamond-core finger-jointed segments into a seamless rigid column of flawless crystal and fibers that can be used to push, pull, or stationkeep, while other nanorobots act to dampen unwanted oscillations.  (Acoustic signals travel 17 kilometers/sec in diamond.)  The entire (reversible) process takes 1-2 seconds.  You need at least three kiloton-test diamond nanohawsers for full positional control.  Each line draws 25 kilowatts of power.


FORCE FIELDS.  To establish a restraining field across a 3 x 3 meter doorway, you release 6 ounces of 0.1-1.0 micron nanorobots along a hairline guide strip on the floor spanning the aperture.  The little buggers climb all over each other, joining hands and raising themselves evenly in a 10-micron-thick multilayered sheet extending from floor to ceiling, wall to wall, following the guide strip, in about 2 seconds.


This barrier won't halt high-speed bullets – you'd need a centimeter-thick nanowall for that.  But the surface is strong enough to stop a human running at world-record sprint speed.  (It can also hold back STP air against vacuum in a 6-meter pressure vessel – makes a dandy airlock or ad hoc spacecraft observation bubble.)  Field nanorobots detect stress and adjust field thickness to redistribute sudden physical or thermal loads.  A reflective surface scatters laser beams harmlessly.  A sharp object like a knife penetrates at the point, but nanorobots grab the blade along the broad flat edge and prevent further entry, then expel the weapon.  The field, roughly 100 nanorobots deep, consumes 1 kilowatt of power to hold structural integrity, can undulate and push back when challenged, and would shimmer, crackle and hum (as in Star Trek) if electrified.


SUPERHERO SUIT.  Imagine wearing a body suit made of intelligent armor fabric half a millimeter thick – probably thinner than the shirt you're wearing now.  Armor fabric is a sheet of 400-micron-thick flawless diamond, broken into small plates to allow flexing and embedded in a diamond mesh fabric.  Plates are coated top and bottom with a 100-micron layer of 1-micron active nanorobots providing structural support, repair, and powerful artificial muscular action.


The suit stops high-velocity (700 m/sec) slugs and 1-ounce shrapnel, and reacts fast enough to stiffen locally under deformation.  If anyone throws a punch, it feels like hitting steel to them, but hardly a tap to you.  The fabric works as a powerful exoskeleton, allowing the wearer to lift (and resist crushing by) up to 5-10 ton loads.  (If you give the nanorobots a couple of seconds to weld the diamond plates into seamless crystal, the garment goes selectively rigid and supports 1000-ton static loads.)  You should also be able to make 15-foot vertical leaps, jog comfortably at 25 mph and sprint to 60 mph.  Of course, the suit does virtually all the work during these feats, piloted, in effect, by its human "passenger."  (Imagine a motorized pogo stick.)  The outfit weighs 4 kg, consumes up to 40 kilowatts of power drawn from 4-kg hour-power packs, and stows itself to the size of a grapefruit.


PHASERS.  Phaser weapons can be nanotechnological devices.  Here's how they might work.  Set to stun, the firearm sprays a narrow low-energy beam of nanorobots capable of inducing coma in the target in under 1 second by entering the body and overloading skin sensor or optic nerve inputs, gobbling neurotransmitters, or scrambling brainstem signals.  Set to freeze, a 1-gram beam of 1-micron nanorobots rapidly encases the target (reversibly) in a skin-tight, steel-rigid webbing, making a living statue much like the "stun gun" in the old Flash Gordon series.


With phaser set to kill, there are countless ways nanorobots can cause near-instant death, some painless, others not.  On Star Trek, the kill setting includes full disintegration.  To kill with disintegration, we spray the target with 0.1-micron nanochoppers studded with arrays of lightweight manipulator arms that can reach out and rapidly yank molecules apart.  Upon firing the weapon, the physical boundaries of the intended target are marked by fast-moving surveyor nanorobots.  A millisecond later, the nanochoppers arrive in large numbers and disassemble all matter within the mapped target volume, then self-destruct.  A 5-gram charge of nanochoppers atomizes a 70 kg human body in 1 second, relying in part on energy released by catalyzing the combustion (in air) of all organics into clouds of water vapor, carbon and nitrogen oxides.


REPLICATORS.  Another Star Trek staple, the food replicator, will put most farmers out of business by the mid-21st century.  Nano-manufactured food will be structurally and nutritionally perfect – no contaminants and precisely specified levels of vitamins, minerals, fats, bones, fibers, sugars, flavors, colors, textures and intoxicants.


As seen on Star Trek, the replicator operates by dropping a cylindrical nanorobot force field onto the worktable, then sterilizing and evacuating the interior.  The force field provides workpiece scaffolding and cooling.  The replicator fills the workspace with the most efficient mix of manufacturing nanomachinery of various sizes.  These nanorobots draw from a raw materials feedstock containing a broad selection of small organic molecules previously extracted from fecal sludge tanks, recycling bins and atmospheric scrubbers.


Let's say you've ordered a juicy 8-oz steak, medium rare.  Following Drexler's "exemplar manufacturing system" (Ref. 10), a 6-kg population of 1017 nanorobots fills a volume of 3000 cubic centimeters and assembles the steak in 4 seconds.  The work volume is flushed clear, the atmosphere replaced.  The force field lifts, and there's your sizzling steak on a plate, ready to eat.  Interestingly, replicated food is far cheaper than farm-grown fare.  Your 8-oz steak costs only 1 megajoule of energy to assemble (3 cents' worth, at today's prices), versus 20 megajoules using cattle, feedlots, slaughterhouses, packers, truckers, and supermarkets.  Even the simplest grain foods cost half as much energy to replicate as to produce agriculturally.  By 2100 AD, farming will be an historical curiosity.


Note that replicators are not elemental transmutators.  They can't turn lead into gold.  All they do is rearrange atoms, already abundantly present, into more useful configurations.


IMMORTALITY.  Your medical exam begins with a comprehensive Cellular Audit.  The doctor sprays you with a fine mist containing a trillion 1-micron diagnostic nanorobots (2 grams) including thousands of functional specialties.  Each of these highly-mobile sensor-studded nanosubmarines (Ref. 18) visits and observes one cell in your body every second, gathering enough data to fill a hundred-page report at each stop.  In 10 seconds, nanorobots visit every cell in your body, report the results to your doctor, and clear out.  Anything suspicious is investigated further and, if necessary, fixed by a second dose of cellular repair machines.


Nanorobots can provide a programmable immune system to rapidly destroy any bacterial, viral or nanomechanical invader listed in the global pathogen database.  Major traumas are easy to repair.  Embedded bullets or shrapnel are exuded, and lacerated tissues repaired, almost instantly.  For example, a 1-inch deep, 1-inch long cut is sterilized and resealed in 1 second using 100 cubic millimeters of active 1-micron nanosuture robots each able to reconnect ten thousand tissue molecules per second.


Senescent cells are reactivated, damaged cells repaired, degenerate brain cells rebuilt, cancerous cells replaced.  Nanodocs should be able to eradicate virtually every cause of aging and death except murder, suicide, legal executions, and accidents involving permanent personality loss.  The 1992 U.S. death rate was 853.3 per 100,000 population, all causes.  Life expectancy was 75.5 years at birth.  Eliminating all but the above four causes leaves just 55.9 deaths per 100,000, an implied life expectancy of 1150 years.  That's effective immortality, by today's standards.


Other biological implications of nanotechnology are equally profound.  We'll soon have to deal with nanoeugenics, bionic pathogens, paleocloning and thanatocloning, man-machine cellular symbiosis, brain transplants, interspecies hybridization, de novo genetic design, biostasis (suspended animation) and ecostasis, nanoterraforming, and – my personal favorite – self-replicating nanorobots.


INVISIBILITY.  Nanotechnology makes possible the Invisible Man.  Imagine a skin-tight suit coated with a layer of digital fisheye optical transceivers.  Photons detected on one side of the suit are measured for frequency and incidence geometry.  This data is relayed to appropriate transceivers on the opposite side of the suit, where photons of like frequency are emitted in such a direction that their line of flight appears uninterrupted.


We have no idea how to engineer this system today, but we do know that nanocomputers can handle the required dataflow.  Here's the worst-case analysis.  Human eye resolution is 1 arcmin.  At closest focus (10 cm), that's 30 microns or 1 billion pixels per square meter.  Assign one fisheye transceiver, and one 1-micron nanocomputer, per pixel.  Each of the 1 billion fisheyes must present one viewplane per square arcmin, a total of 100 million viewplanes.  Using 20-bit words to code 1 million distinct color values and scanning at 100 frames/second, we process 2 x 1020 bits/sec per square meter of suit surface.  This computational load is easily handled by 2 milligrams of nanorobots consuming 1 kW of power, backed by a densely webbed optical fiber databus.


We can use the same technology to camouflage or "cloak" ground vehicles, buildings, airplanes and spacecraft.


TRANSPORTERS.  General-purpose matter transmission appears possible using nanosystems, but really pushes the edge of the envelope.  The crude method I'll describe has three serious drawbacks.  Transmission must be:  (1) line of sight (no beaming through walls),  (2) station-to-station (must have transmitter and receiver), and (3) in vacuo (no beaming through atmospheres).


To transmit, a 100-kg payload is placed on a 1-meter diameter floor pad.  The transmitter system throws a restraining net over the payload to immobilize it for scanning, then drops a cylindrical force field the diameter of the pad and evacuates the cylinder.  Eleven tons of nanorobots of various sizes fill the cylinder and perform a hierarchical decomposition of the payload using a fixed scan geometry, reading each of 1028 atoms as an 8-bit word thus generating 1029 bits of structured compositional data.  Using a 5 gigawatt nanolaser array and one blue photon per bit, we transmit the entire description in 10 seconds – consuming 50 gigajoules of energy (about $1500 worth at today's power costs) – to a distant receiver, say, 100 kilometers away, where the compositional data is decoded, driving payload reconstruction.


It ought to work fine with inanimate objects, but would live cargo survive?  I'm with Dr. McCoy – you couldn't pay me to step into the thing!


THE HOLODECK.  Nanorobot populations can configure themselves into complex surfaces of any shape, size, or texture.  You could carry a 1 kilogram of them – enough to construct 50 square meters of programmable 10-micron surface – in your handheld MakePad, a combination control and storage unit hung at the belt.  These nanorobots could be programmed to make force fields in the shapes of chairs, tables, beds, even a smart trampoline.  (Wouldn't that be fun!)


For the holodeck, more elaborate nanosurfaces are configured into shapes resembling people, animals, plants, buildings or machines, using mostly ballast mass and structured filler material to build to minimum human eye resolution.  Hefty onboard computer power plus artificial internal tendons, bellows, heaters and other specialized nano-organs give these assemblages the ability to move around, talk, eat, and otherwise perfectly simulate the real thing.


NANOGHOSTS.  An aerogel of nanorobots would make an excellent poltergeist!  A structured cloud of 1-micron nanorobots weighing 50 milligrams occupying the volume of a human body could retain its physical integrity in a 60 mph wind, lift and hurl 1-kg objects, emulate an angry personality, suck energy out of electrical outlets, and reshape itself into jets & airfoils to flap around rooms riding the 0.1-0.5 mph household currents caused by fans, air conditioners and human movement.


PRECOGNITION AND PSYCHOHISTORY.  The advent of nanocomputers will make it possible to create detailed simulations of human sociocultural process.  Elaborate psychosocial models of individual people, their communities and physical environments might lead to foreknowledge of general outcomes in personal lives about as reliable as 1-2 week weather forecasts are today.


Scientists in a single cliology laboratory could assemble 15 billion Human-Equivalent Computers (one for each person on Earth in the year 2040), a sugarcube-size metacomputer weighing 2 grams.  Researchers could then run historical simulations of human civilization a million times faster than real time, testing theories and developing reliable forecasts.  Human cultural evolution will become at least as predictable as the behavior of social insects is today, although mechanocultural evolution may remain intractable.


PSIONICS, MAGIC AND MIRACLES.  By positioning one telemetered nanorobot in each of the brain's 10 billion cells (total of 20 milligrams, 1 watt of waste heat), it becomes possible to directly read and transmit real-time brain states from one mind to another – true telepathy.  Once a mental link exists, you could use it to command the activity of nanorobots around you to move objects (telekinesis), to harm others (hexing), burn objects (pyrokinesis), to mindread other people (empathy) or animals (telezoopathy), to sense remotely (dowsing), to teleoperate or "waldo" another brain (zombification), or to mind-clone (possession).  Nanotechnology will also make it easy to wow the primitives (anyone lacking post-20th-century knowledge) with magic tricks or miracles such as levitation, patterned bleeding (stigmata), turning staffs into serpents and (muddy river) water into wine, healing the sick, parting seas, and so forth.






A 2-square-meter nanosurface (human-size) comprised of 4-micron nanocomputers, 32 microns thick, represents a population of a trillion individual nanorobots each possessing teraflop (human-equivalent) intelligence but collectively weighing only 1 pound.  By the mid-21st century, we may see a startling variety of strangely-shaped organisms made up of trillions or quadrillions of independent HEC nanocells, all voluntarily associated on a temporary basis to achieve some (to us) mysterious objective.  Even the smallest of these macroscopic creatures will be protean in shape and mass, something like the liquid robot in Terminator II, but hypersentient.  By comparison, modern conceptions of advanced robots like Star Trek's Mr. Data appear laughingly primitive.


Consider:  Each nanoorganismic collective is a microscopic nation of intelligences more akin to human civilization than to individual people.  A nanocreature could bud off tiny chunks of itself, each capable of independent motion with superhuman mentality, to serve as information seekers, materials gatherers, weapons, or other independent agents.  Like the cartoonish Power Rangers, nanoorganisms could temporarily merge to form larger entities for special purposes.  Nanocreatures could manufacture virtually anything and reproduce at will, qualifying by any reasonable measure as a higher form of life than our own.


How would they regard us?  Mathematician I.J. Good once remarked that "if we build an ultraintelligent machine, we will be playing with fire.  But we have played with fire before, and it has helped to keep the other animals at bay."  Hearing this, Arthur C. Clarke retorted:  "When artificial intelligence does arrive, we shall be the other animals;  and look at what has happened to them.  It will be poetic justice."


So get ready.  It's going to be an interesting century.






1.  Hans Moravec, Mind Children:  The Future of Robot and Human Intelligence, Harvard University Press, Cambridge, MA, 1988.


2.  Mark Johnson, "Bipolar Spin Switch," Science 260 (16 April 1993):320-323.


3.  Faye Flam, "Researchers Defy the Physical Limits to Computation," Science 260 (16 April 1993):290-292.


4.  S. Hasuo, "Toward the Realization of a Josephson Computer," Science 255 (17 January 1992):301-305.


5.  M. Mitchell Waldrop, "Computing at the Speed of Light," Science 259 (22 January 1993):456.


6.  Seth Lloyd, "A Potentially Realizable Quantum Computer," Science 261 (17 September 1993):1569-1571 and 263 (4 February 1994):695.


7.  John Travis, "Building a Baby Brain in a Robot," Science 264 (20 May 1994):1080-1082.


8.  K. Eric Drexler, Engines of Creation:  The Coming Era of Nanotechnology, Anchor Press/Doubleday, New York, 1986.


9.  K. Eric Drexler, Chris Peterson, Gayle Pergamit, Unbounding the Future:  The Nanotechnology Revolution, William Morrow/Quill Books, NY, 1991.


10.  K. Eric Drexler, Nanosystems:  Molecular Machinery, Manufacturing, and Computation, John Wiley & Sons, NY, 1992.


11.  Nanocon Proceedings, Seattle, Washington, 17-19 February 1989, July 1989, NANOCON, Box 40176, Bellevue, WA 98004.


12.  B.C. Crandall, James Lewis, eds., Nanotechnology:  Research and Perspectives, MIT Press, 1992.


13.  Richard Terra, "Working Molecules:  Progress Toward Nanotechnology," Analog 113 (October 1993):44-68.


14.  Proceedings of the Third Foresight Conference on Molecular Nanotechnology, 14-16 October 1993, Palo Alto, CA; in Nanotechnology 3 (July 1994).


15.  B.C. Crandall, ed., Nanotechnology, MIT Press, 1995.  In preparation.


16.  Yun Kim, Charles M. Lieber, "Machining Oxide Thin Films with an Atomic Force Microscope:  Pattern and Object Formation on the Nanometer Scale,"  Science 257 (17 July 1992):375-7.


17.  T. Ross Kelly, Chemical & Engineering News, 25 April 1994, pp. 6-7.


18.  A. K. Dewdney, "Nanotechnology – Wherein Molecular Computers Control Tiny Circulatory Submarines," Scientific American 258 (January 1988):100-103.