Jonathan V. PostCEO: Computer Futures, Inc.
3225 N. Marengo Ave.
Altadena, CA 91001
Dr. Donald D. RoseCEO: Creativision Consulting
1033 Hilgard Ave., #214
Westwood, CA 90024
While Artificial Intelligence (AI) has become increasingly present in recent space applications, new missions being planned will require even more incorporation of AI techniques. In this paper, we survey some of the progress made to date in implementing such programs, some current directions and issues, and speculate about the future of AI in space scenarios. We also provide examples of how thinkers from the realm of science fiction have envisioned AI’s role in various aspects of space exploration.
Dedicated to Dr. Isaac Asimov: scientist, writer, visionary, friend
1. Introduction One morning as I was leaving for my space program job, my two-year-old son Andrew
Carmichael Post pleaded with me to stay home. "Don't go to work, Daddy" he said,
hitting me with a coherent beam of cutons, the quanta of cuteness, "stay home and
play with me all day."
"I wish I could, Darling, but I need to go to work."
"Because I need to get some money."
"But I have money for you Daddy, wait..." and he ran to his room, rummaged
around, and returned. "Here, I have have two monies. You can have them.
Now stay home with me."
I thanked him for the two pennies, and tried again. "Andrew, I must go to work to help
build rockets and robots."
"Rockets and robots are my two favorite toys. We could make rockets and robots
I did finally get away to work, not without regrets, and just realized as I
rewrote this column that its subject is nearly identical to my son's enthusiasm.
Rockets and robots or, more broadly, space exploration and artificial intelligence.
As humans venture further away from Earth, the need for autonomous systems, and
hence capabilities developed for Artificial Intelligence (AI), will increase dramatically.
The increased danger inherent in longer duration missions, among other reasons, will
make the role of AI essential -- e.g., to avoid or minimize the need for humans
on such missions, and to augment the abilities of those humans still present.
In this article -- the final science column for Quantum and the only one co-authored
with Dr. Donald David Rose -- we discuss some of the applications already developed
for space applications, then venture further into the hypothetical future to discuss
how various domains of space investigation might benefit from AI.
Along the way, after
(1) this introduction, we will discuss:
(2) the control and diagnosis of space-related environments by expert systems;
(3) Traversing Extraterrestrial Sites with Telerobotics and Virtual Reality,
dumbots and replibots;
(4) Artificial Intelligence for Scientific Investigation, including data analysis
(5) a survey of Science Fiction Definitions and Science Fictional Space Travel
(including androids, waldos, terraforming, faster-than-light, and space opera);
(6) a survey of Science Fictional Views of Computers and Robots (including cyborgs);
(7) Science Fictional Views of Artificial Intelligence; and
(8) a wrap-up and conclusion section, just down the hall from the holodeck.
There used to be a Reference section to this column, which drew heavily on [Mike Ashley's
The Illustrated Book of Science Fiction Lists, NY: Cornerstone (Simon & Schuster),
1982], and [Jakubowski, Maxim and Edwards, Malcolm, The SF Book of Lists, NY:
Berkley, November 1983] but these references have been chopped up and sprinkled through
the text -- it might slow down the plot, but it takes that dry academic taste out of your
Fire up your rockets and power up your robots, here we go!
2. Environment Maintenance
Several AI systems have been successfully developed and deployed to control and/or
diagnose space-related environments, e.g., to ensure that hardware and software are
performing within desired parameters, and finding the cause of faults when they occur. For
example, diagnosis as well as control of hardware and software has also been proven in expert
systems, such as g2 (developed by Gensym Corp.). The latter has been used to control and
diagnose environments such as BioSphere2 (BS2) and the Space Shuttle. BioSphere2, in
particular, provides an excellent testing ground for AI systems that will soon be needed in space,
especially Lunar Base applications, since BS2 is an enclosed self-sufficient habitat that happens
to be on Earth. Hence, knowledge learned by applying AI code in BS2 should be almost directly
applicable to upcoming Lunar Base missions, and eventually Mars habitats as well.
Whereas the above examples focus on AI applied to monitoring and diagnosis of entire
environments involving several interacting systems, researchers have also developed and
deployed many subsystem-specific programs for device and vehicle diagnosis. In
fact, these make up some of the most common and successful applications of AI to space
domains to date. For example, NASA has used a system called PI-IN-A-BOX to automate the
diagnosis of equipment failure about the Space Shuttle. Another NASA project has been
automating the diagnosis of a specific Shuttle subsystem (the Reaction Control System). This
latter system is currently being tested on the ground, and will probably evolve into a system for
use by ground-based mission controllers -- but a later more advanced version of this RCS
diagnostician could wind up on the Shuttle or its descendant craft. Rockwell International has also
been constructing expert systems, to diagnose other parts of the Shuttle such as its fuel cell and
In summary, AI has already made valuable contributions to this field, and increases in flight
duration and craft complexity (e.g., Space Station Freedom with its 30 year life span, as well as
Lunar vehicles and habitats designed for continuous long-duration use) will make sophisticated
"artificial diagnosticians" even more essential. In particular, programs will have to help unmanned
vehicles repair themselves if needed, such as during unmanned rover "field trips".
Current machine learning (ML) techniques, which have already been used to
augment diagnosis systems, should help in this regard. For example, ML can be used to help
predict impending faults before they occur, so that system disruptions and downtime can be
minimized. (Fans of the film 2001 [Clarke, Arthur C., 2001: A Space Odyssey, NY: NAL, 1968;
also see sequels (2010; 2061)] might note that the HAL 9000 computer exhibited such a
capability, when it informed its crewmates that an earth-acquisition antenna was about to fail.) In
general, the use of AI technology for diagnosis will greatly decrease the need for EVAs that are
not related to purely scientific objectives.
3. Traversing Extraterrestrial Sites
Assuming that one’s hardware and software are functioning normally, one of the most
important tasks to do next is exploration. This includes deciding which sites should be
investigated -- e.g., deciding which areas have greatest potential for scientific results and
estimating the likely danger in getting there -- and then actually traversing to the sites of interest. Telerobotics (TR) could be used to decrease the number of human EVAs required on
space missions. In fact, Marvin Minsky (a father of both AI and TR) observed that, if we had
thought ahead, we could have had an inexpensive teleoperated rover doing meaningful traverses
on the moon for the two decades that have elapsed since humans last left a footprint on our
natural satellite. Since this indeed seems a cost-effective option, especially in relation to other
proposed missions, we should certainly reexamine its use today as attention is refocused on the
TR is part of the larger field of telepresence, itself a subarea of Virtual Reality (VR).
The idea here is to put a human "virtually" in some dangerous locale (e.g., space) via one or more
robots, which provide the virtual eyes and perform tasks with virtual limbs as the human remains in
a safe haven. Intelligent software will be needed to develop effective telepresence capabilities -
mainly via ground control in the near-term, but applicable in several space-based domains in the
not-too-distant future. For instance, to decrease human risks in constructing and maintaining
Space Station Freedom or remote interplanetary bases, a human could operate a robotic repair
droid remotely from inside the Station, or from inside a habitat if the locale is a remote lunar or
Martian site. Also note that, for relatively “terraclose” applications (ranging from Earth to locations
near the Moon) one could even perform teleoperations from Earth itself without appreciable lag.
As far as longer-term longer-range robotics applications are concerned, one of the most
exciting will be the return of our presence to the Moon and Mars. A major player in the latter
domain is JPL, which has been developing intelligent robots to roam the Martian surface or
other extraterrestrial sites in a way that can maximize scientific gain while minimizing time and
danger to a mission. Looking further ahead, several intriguing and challenging issues are
unfolding for robotic space domains. One general issue involves deciding on the best approach
to designing and deploying robots, for both nearby and remote Voyager-like space missions.
One of the most common ideas has been to build one or a few robots with complex, intelligent
processing onboard each one.
However, an alternate approach, favored by researchers such as MIT's Rodney Brooks,
would use of tens or even hundreds of smaller, less (individually) intelligent robots per mission.
His slogan is "small, fast, and out of control." One advantage of using an army of "dumbots"
would be increased fault-tolerance. If you lose one robot, many others remain to finish the
mission goals. Another approach would use one or more self-replicating robots. In fact, this latter
idea might lead to a compromise between the first two approaches; one could deploy a small
number of intelligent "replibots" which would create small dumbots during a mission, even
replace those that fail as needed. The longer the duration of a mission, such as an extended stay
at a lunar or Martian base, the more important replication or automated droid
NASA likes "small", NASA likes "fast", but NASA can't stand the risk of "out of control."
This reaises the question of whether space exploration ought to be under the control of risk-
averse government agencies, or risk-tolerent free enterprise corporations. Once the corporate
world has "replibots", the balance will change forever. Nanotechnology is another path to
replication, and a possible short-cut to planetary exploration, but that's a subject (however
dear to my heart) outside the scope of this column.
Maybe we go to space for political reasons, maybe we go for psychological reasons,
maybe we go for profit, but as far as the academic world is concerned, the only reason that
matters is science. Which leads us to the next topic....
4. Scientific Investigation
Once a human or machine is actually at a potentially valuable site, the priorities for AI
applications shift to actual scientific tasks, such as chemical analyses, deciding which samples to
carry back to a habitat or spacecraft, etc.
One fruitful area of AI that applies to these and other scientific tasks is the field of data
analysis and discovery. Several programs have been developed in this category. For
example, NASA scientists have constructed and are refining the AUTOCLASS system, which
automatically classifies data into meaningful groups. Its use to date has included classifying
well- known star classes, as well as discovering new astronomical classes (e.g., separating data into
classes having only very subtle differences in spectra). Such systems not only can lead to new
knowledge, as just described, but will become essential for sifting through the mounting
quantities of mission data being gathered from space missions.
AUTOCLASS and other AI systems should aid efforts being made in the general area of
"ground-based space exploration", including projects such as SETI (the Search for
Extraterrestrial Intelligence), which begins a vastly enhanced search in Fall 1992. Sifting through
mounds of extraterrestrial data and noise for meaningful intelligent signals seems like an ideal
challenge for AUTOCLASS or related systems.
But what are the criteria for "meaningful signals?" The most efficient communications,
according to Claude Shannon's theory of channel capacity [Shannon, Claude,
Communications Theory, MIT Press, 1945] are statistically indistinguishable from noise. It looks
like a noisy universe out there -- but maybe that's a million interstellar civilizations just to save on
the galactic phone bill! Which leads us right into the topic of Science Fiction....
5. Past Looks into the Future: Science Fiction Definitions and Science Fictional Space Travel
Science Fiction (SF) preceded science fact in every major qualitative aspect of computers,
artificial intelligence, robotics, and space flight. The more than 140 citations scattered through the
text from the extensive former reference sectio, represents the tip of the iceberg. Science fiction
is, among other things, a visionary literature that has many definitions. Among these, and relevant
to this paper: "Science fiction is a branch of fantasy identifiable by the fact that it eases the 'willing suspension of disbelief' on the part of its readers by utilizing an atmosphere of scientific credibility for its imaginative speculations in physical science, space, time, social science, and philosophy." [Moskowitz, Sam, Explorers of the Infinite, 1963] "We might try to define science fiction in this broader sense as fiction based upon scientific or pseudo-scientific assumptions (space-travel, robots, telepathy, earthly immortality, and so forth) or laid in any patently unreal though non-supernatural setting (the future, or another world, and so forth.) [de Camp, L. Sprague, Science Fiction Handbook, 1953] "Science fiction is that branch of literature that deals with human responses to changes in the level of science and technology." [Asimov, Isaac, Isaac Asimov's Science Fiction Magazine, March-April 1978] The legacy of Isaac Asimov, one of SF’s best known progeny (who passed on as this paper was being prepared), includes many concepts that influenced AI, such as the invention of the word robotics, and the famous “Three Laws” that he (followed by countless others) felt should govern their use. Asimov also influenced generations of scientists and technologists, such as Marvin Minsky, himself one of the top experts in robotics and a pioneer of both AI and telepresence. Science Fiction, of course, supplied the original word robot, from the Czech "robota" (laborer), in a stage play [Capek, Karel, R.U.R. (Rossum's Universal Robots), play opened in Prague, 26 January 1921]. Science Fiction provides [Williamson, Jack, "The Cometeers", Astounding, May 1936] the related term android (hence the contraction droid), meaning a robot with a humaoid appearence or, more commonly, an artificially created organic humanoid. Science Fiction also predicted teleoperation, with the word waldo [Heinlein, Robert A., "Waldo", 1942], which was adopted when the technology came into existence later. Even the most sub-literary "pulp" science fiction absorbed the spirit of scientific/engineering methodology: "If you could translate a problem into computer terms, then there was no problem that you couldn't answer. It all depended on a man's ability to translate. To get the feel of a problem. To sense it. To view the whole cosmos as a huge mathematical equation." [Fanthorpe, R.L. (Robert Lionel), The Asteroid Man, Badger: SF35, 1960]. Science fiction created, over three centuries of speculation, the enthusiasm with which the real space program was born. Three centuries? Yes, and we begin by very briefly considering 15 very early and very unusual books about space voyages [Ashley, Mike, The Illustrated Book of Science Fiction Lists, NY: Cornerstone (Simon & Schuster), 1982]. In the two following sections (6. Science Fictional Views of Computers and Robots; and 7. Science Fictional Views of Artificial Intelligence) we relate space travel with AI. A Voyage to the World of Cartesius [Daniel, Gabriel, A Voyage to the World of Cartesius, 1692] fictionally explores the cosmos of rotating vortices hypothesized by Descartes circa 1640, a cosmos recently confirmed by studies of the origins of the solar system, of spiral galaxies, and of accretion disks around black holes. The History of Israel Jobson [Wilson, Miles, 1757] starts by climbing a ladder to the moon from Pen-Y-Ghent, Wales. This mountain now is a major center for hang-gliding. And "ladders", "beanstalks", "skyhooks" and "space elevators" and other words now apply to the notion of tethers which have indeed been proven feasible in principle to dip down to planetary surfaces and carry payloads into orbit with no expenditure of fuel [Clarke, Arthur C., The Fountains of Paradise, NY: HBJ, 1978];[Sheffield, Charles, The Web Between the Worlds, NY: Ace, 1979]. The protagonist then takes the first fictional tour of the solar system and beyond, albeit in a version of Elijah's chariot. The first fictional voyage to Uranus ["Vivenair", A Journey Lately Performed Through the Air, in an Aerostatic Globe, to the Newly Discovered Planet, Georgium Sidum, 1784] was a source of inspiration to one of this paper's authors (JVP), who years later worked as Mission Planning Engineer for the Voyager spacecraft's flyby of the actual planet Uranus, and who was primarily responsible for the innovative and very successful design of the flyby and imaging of the Uranian moon Miranda -- itself named after a fictional character. In [Erskine, Thomas, Armata, 1816] an imaginary planet attached to the South Pole, and reached by boat, prefigures the reality of dumbell-shaped double asteroids and the astrophysical possibility of pairs of planets close enough to share atmospheres or even oceans [Forward, Robert, Rocheworld (and sequels) ]. The first genuinely alien lifeforms in any writings appear in [Peabody, Joel R., A World of Wonders, 1838], in a mental tour of the solar system. As rock group The Moody Blues puts it ["In Search of the Lost Chord"] -- "seen through the universe, thinking is the best way to travel..." Rocket propulsion is used to disasterous effect in the earliest science fiction [de Bergerac, Cyrano, Voyage to the Moon, 1650] and more effectively in [Perce, Elbert, Gulliver Joi, 1852], in which the eponymous hero travels by rocket-propelled spaceship to the newly discovered planet Kailoo. In a still-fanciful far-future technology, matter transmission ("beam me up, Scotty") first appears in [Whiting, Sidney, Helionde, 1855] as a way of reaching an inhabited Sun. Another early use of matter transmission is The Man Without a Body [Mitchell, Edward Paige, The Man Without a Body, in newspaper: New York Sun, 25 March 1877] and, in a fascinatingly more technical view in [Jane, Fred T., To Venus in Five Seconds, 1897]. Since a few grams of matter contains quintillions to sextillions of atoms, the matter transmission problem is that awful bottleneck, the computer. Computer systems big enough and fast enough to encode every atom in a person have problems of their own, including quantum noise effects. And any system that can beam up a person can also duplicate a person. As several SF writers have pointed out as well, would you let yourself be killed by such a system just because a perfect copy of you would be replicated somewhere else? The insurance companies would never allow it.... Manned Earth satellites were predicted in detail a century before actualized, in [Hale, Edward Everett, "The Brick Moon", Atlantic Monthly, October 1864, and in The Brick Moon and Other Stories, Boston: Little Brown, 1899]. Here, a 200-foot ceramic satellite is launched by flywheel. This also forecasted the use of ceramics in space -- as in the Space Shuttles' thermal tiles, and the possibility of flywheel launch (recently re-examined for Moon-based applications by Dr. Robert Waldron). "Vulcan" was a planet hypothesized by Leverrier in 1845 to explain anomalies in the orbit of Mercury. Vulcan supposedly orbited the sun entirely within the orbit of Mercury. It was fictionally visited in ["Nunsowe Green", A Thousand Years Hence, 1882]. The word is more commonly used today as the name of Science Officer Spock's home planet. Anomalous precession of Mercury's perihelion was first correctly explained by Albert Einstein. The first written treatment of both terraforming and faster-than-light space travel are in the same book: Man Abroad [Anon 1886]. Here, planetary colonization is taken for granted, on planets artificially modified to be habitable to humans. This book is also the first space opera: a sub-genre in which there is epic conflict between the inhabitants of more than one planet. Galle discovered the planet Neptune in 1846, as predicted by astronomers Leverrier and Adams. It's first fictional visit -- glacial but inhabited -- was in ["Spirito Gentil", Earth-Born!, 1889]. The idea of a counter-Earth, in our same orbit around the Sun, but always 180 degrees ahead of us in that orbit, and hence invisibly on the other side of the Sun, is in [Stump, D.L., From World to World, 1896, revised as The Love of Meltha Laone, 1913]. The same setting is used in the notorious sexist and sadistic Gor novels [Norman, John -- pseudonym] and a recent pastiche [Lupoff, Richard A., Contraterran]. In [Parmele, Mary Platt, Ariel, 1898] "the power of the authorial mind has created an artificial satellite 400 miles across and 400,000 miles distant." Frankenstein's monster lives there, as do other imaginary characters [Locke, George, Voyages in Space, valuable in compiling these references]. Interstellar conflict is first fictionalized in [Cole, Robert W., The Struggle for Conflict, 1900], when in 2236 the "Anglo-Saxon Federation" has colonized the solar system, but is on the brink of war with the advanced race of the planet Kailoo -- lifted from [Perce, Elbert, Gulliver Joi, 1852] which circles the star Sirius. Percival Lowell predicted, circa 1890, the existence of a planet beyond Neptune. Clyde Tombaugh discovered that planet in 1930, and named it Pluto , a name appropriate both for the god of the frozen wasteland of hell, and whose first two letters stand for Percival Lowell. [Horner, Donald W., Their Winged Destiny, 1912] described spaceflight to Alpha Centauri by astronauts who, as they leave the solar system, pass a planet beyond Neptune. [Coblentz, Stanton, In Plutonian Depths, in Wonder Stories Quarterly, Spring 1931; NY:Avon, 1950] quickly takes advantage of the newly discovered and named world. The Disney dog was named for the planet, during animations ongoing in 1930. Little kids have been confused ever since. In these examples, we see that spaceflight has been a frequent subject for novels far earlier than the 20th century in which technology made spaceflight possible. In the next section, we see the same pattern in fictional forecasts of computers and robots. 6. Science Fictional Views of Computers and Robots Although the conventional belief is that literature is about people, science fiction has expanded the role of literature to feature computers, especially artificially intelligent computers, as personalities worthy of story-telling. Just to mention 15 quick examples of computers and AI computers with proper names [Jakubowski, Maxim and Edwards, Malcolm, The SF Book of Lists, NY:Berkley, November 1983, another valuable aide in compiling these references] who appear in stories and books, we have: Abel [Durrell, Lawrence, Tunc and Nunquam], Bossy [Clifton, Mark and Riley, Frank, They'd Rather Be Right, a.k.a. The Forever Machine, NY: Gnome, 1957; Galaxy SF Novel 35, 1959, based on Clifton's original "Joey" stories], Colossus [Jones, D.F., Colossus, London: Hart- Davis, 1966; NY:Putnam, 1967], Domino [Budrys, Algis J., Michaelmas, NY:Berkley, July 1978], Epicac XIV [Vonnegut, Kurt, Player Piano, NY:Scribner, 1952], Epikistes [Lafferty, R.A., Arrive at Eastermine], Extro [Bester, Alfred, The Computer Connection, Conde-Nast, 1974, 1975], HAL 9000 [Clarke, Arthur C., 2001: A Space Odyssey, NY: NAL, 1968; also see sequels (2010; 2061)], Harlie [Gerrold, David, When Harlie Was One, NY:Ballentine, 1972], the "fair dinkum thinkum" Mike [Heinlein, Robert A., The Moon is a Harsh Mistress, NY:Putnam, 1966], Multivac [Asimov], Proteus [Koontz, Dean R., Demon Seed], Shalmaneser [Brunner, John, Stand On Zanzibar, NY:Doubleday, 1968], Tench 889B [Dick], and Unicomp [Levin, Ira, This Perfect Day]. There has even been an opera about an AI system [Johannesson, Olaf (pseudonym of Swedish Nobel laureate cosmologist Hans Alfven), The Tale of the Big Computer, Stockholm: Albert Bonniers, 1966; tr. Naomi Walford, Coward-McCann, 1968; London: Gollancz, 1968; Royal Swedish Opera, 1967], in which the future evolution of life on Earth moves on beyond humans to the "Symbiotic Age" and then to the start of the true age of computers. Sometimes computer intelligence is imagined to lead to mad, megalomaniac, or godlike capabilities. In [Brown, Frederic, "The Answer",] a computer is asked if there is a God -- and answers "There is ... now!" This story has passed into folklore, and is told by thousands of people who never heard of the actual author. In "The Last Question" [Asimov], a computer circles back in time to create our universe, or perhaps the next universe after this one. [Ellison, Harlan, "I Have No Mouth and I Must Scream", ] is a terrifyingly realized tale of a vindictive computer torturing the last remaining humans after destroying civilization -- surely the blackest picture ever painted of AI. In two related novels, [Herbert, Frank, Destination Void, NY:Berkley, 1966; Herbert, Frank & Ransom, Bill, The Jesus Incident, NY: Berkley-Putnam, 1979], AI systems in space evolve to trans-human and theological power. In [Silverberg, Robert, "Going Down Smooth"] a computer slips into psychosis, but does a credible job as a psychiatrist meanwhile a profession that certainly might prove valuable on long missions if physical human presence continues in space. In the chillingly subtle [Williamson, Jack, The Humanoids, NY:Simon&Schuster, 1949] AI robots do such a good job of freeing people from risk and labor that we become pampered pets of the computers -- but never to shake off the leash. In the best "space opera" in years, [Banks, Iain, Consider Phlebas, London: MacMillan, 1987], the space AI systems of the largest spacecraft (General Contact Units) are called "Minds", and are far more intelligent than people: "The War had to be part of the Minds' idea; it was part of their clinical drive to clean up the galaxy, to make it run on nice, efficient lines, without waste, injustice, or suffering. The fools in the Culture couldn't see that one day the Minds would start thinking how wasteful and inefficient the humans in the Culture themselves were." More recently [Vinge, Vernon, A Fire Upon the Deep, NY: Tor, April 1992] it has been suggested that anisotropic physical constraints prevent "true" AI from implementation at Earth's distance from the center of the galaxy (the Great Slowness) but that further out (the Beyond) AI becomes easier, and further out still (the Transcend) godlike AI is simple -- and sometimes dangerous -- to develop. Although we have seen AI computers as fictional personailities, science fiction admits to a favoritism for intelligent robots as such. A dozen examples of named robots are provided: Adam Link [Binder, Eando, Adam Link, Robot ;(made into Outer Limits episode)], Brillo [Bova, Ben, and Ellison, Harlan, "Brillo"; Future Cop, Taylor, Jud, made-for-TV movie (resulted in $285,000 settlement paid to Ellison/Bova ], Helen O'Loy [del Rey, Lester, "Helen O'Loy", in ... And Some Were Human, Philadelphia: Prime, 1948], Jasperodus [Bayley, Barrington J., The Soul of the Robot,1974], Jay Score [Russell, Eric Frank, Men, Martians, and Machines, UK: Dobson, 1955, NY:Roy, 1956], Jenkins [Simak, Clifford, City, series in Astounding Science Fiction: May 1944, July 1944, September 1944, November 1944, January 1946, November 1946, December 1947], Krag [Hamilton, Edmond, Captain Future series, 1940 ff], Marvin [Adams, Douglas, Hitchhikers Guide to the Galaxy, see film Hitchhikers Guide to the Galaxy, 1980], R. Daneel Olivaw [Asimov, Isaac, The Caves of Steel, NY:Doubleday, 1954; Asimov, Isaac, The Naked Sun, NY:Doubleday, 1957], Roderick [Sladek, John, Roderick and Roderick at Random], Spofforth [Tevis, Walter, Mockingbird, also author of The Man Who Fell to Earth], and Tik-Tok [Baum, L. Frank, Oz series, beginning with The Wonderful Wizard of Oz, 1900; see also Jack Snow, Who's Who in Oz, 1954]. Robots are attractive in more ways than one. Many stories [Jakubowski, Maxim and Edwards, Malcolm, The SF Book of Lists, NY:Berkley, November 1983] deal with sex between humans and robots. Perhaps this is the only truly safe sex ("Applied Dildonics" vs. "Meat Sex") possible in the otherwise infectious organic world. Such examples include [Bloch, Robert, "The Tin You Love to Touch", Other Worlds, June-July 1951], [del Rey, Lester, "Helen O'Loy", in ... And Some Were Human, Philadelphia: Prime, 1948], [del Rey, Lester, "A Pound of Cure", New Worlds #27, September 1954], [Groves, J.W., "Robots Don't Bleed"], [Leiber, Fritz, "The Mechanical Bride"], "The Stepford Wives" [Levin], [McIntosh, J.T., "Made in U.S.A.", Galaxy, April 1953], [McLeod, Shiela, Xanthe and the Robots], [Rotsler, William, "Ship Me Tomorrow"], [Saxton, Josephine, "Gordon's Women"], [Sheckley, Robert, "Can You Feel Anything When I Do This?"], [Silverberg, Robert, Tower of Glass], [Wilhelm, Kate, "Andover and the Android"], and [Young, Robert F., "September Had Thirty Days"]. Robots with distinct names and personalities make frequent appearences in movies, beginning with "The Clown and the Automaton" [Melies 1897], and including: C3P0 and R2D2 [Star Wars 1977], Maria [Metropolis 1972], Robby [Forbidden Planet 1956 and The Invisible Boy 1957], Gort [The Day the Earth Stood Still 1951], Huey, Dewey, and Louie [Silent Running 1972], The Cyclons [Battlestar Galactica 1978], The Ice Robot [Logan's Run 1976], Gunslinger (Yul Brenner) [Westworld 1973], The Wives [The Stepford Wives 1975], Woody Allan as Miles Monroe impersonating robot servent [Sleeper 1973], Twiki and Dr. Theopolis [Buck Rogers in the 25th Century 1979], Marvin the Paranoid Android [TV version of Hitchhikers Guide to the Galaxy 1980], Tobor the Great [?], Bubbo [Clash of the Titans 1981], Vincent [The Black Hole 1979], the smash hits Terminator and Terminator2, and in made-for-TV movies such as [The Questor Tapes 1974] and [Future Cop]. Another way that we become close to our machines is by merging with them to become cyborgs -- cybernetic organisms -- a term invented by science fiction [Caidin, Martin, Cyborg, basis for TV series The Six Million Dollar Man] and now in general use. Fictional examples [Jakubowski 1983] include: [Bayley, Barrington J., The Garments of Caean], [Budrys, Algis J., Who?, NY:Pyramid, 1958], Cyborg [Caidin], [Compton, D.G., The Continuous Katherine Mortenhoe, a.k.a. The Unsleeping Eye], [Delany, Samuel R., Nova, NY:Doubleday, 1968], [Knight, Damon, "Masks"], [McCaffrey, Anne, "The Ship Who Sang", Fantasy & Science Fiction, April 1961], [Moore, C.L., "No Woman Born"], [Smith, Cordwainer (pseudonym of Paul Myron Anthony Linebarger), "Scanners Live in Vain", in You'll Never Be the Same, Regency, 1963], and the Dumarest series [Tubbs, E.C. [Edwin Charles], starting with The Winds of Gath, Ace, 1967]. In particular, The Ship Who Sang and "Scanners Live in Vain" posit the need to mechanically alter humans profoundly in order to make long-duration spaceflight feasible -- a worst-case analysis (from one point of view) if the problems of weightlessness cannot be overcome by more conventional means But see "Loading Up For Liftoff", Jonathan V. Post, Ad Astra, March 1992, for evidence that all it takes to avoid the biological hazards of zero gravity is to be overweight, aerobically unfit, and have high blood pressure! Another possible advantage of having humans in more direct, detailed contact with machines is greater control over how the software’s knowledge evolves -- such as keeping human goals, like relevance and explainability, paramount over those the AI software might have. The human being is either master or slave to the machine ... which would YOU rather be? 7. Science Fictional Views of Artificial Intelligence AI was forecast by science fiction over a century ago. In The Ablest Man in the World [Mitchell, Edward Paige, The Ablest Man in the World, 1879], a calculating machine is transplanted into the brian of an idiot. He becomes a super-genius. Similarly, [Mitchell, Edward Paige, The Tachypomp, 1874] forecasts significant social upheavals from the creation of a supercalculator. More recently, at least three science fiction authors simultaneously have considered the implications of Charles Babbage having succeeded with the Difference Engine and the Analytical Engine, thereby bringing about the computer age a century ahead of time [Gibson, William and Sterling, Bruce, The Difference Engine, NY: Bantam 1991], [Post, Jonathan V., Headcrash, novel outline to Analog, 1988], [Sheffield, Charles, “Georgia on my Mind”, 1993 (in press)]. In literary roots, however, AI and robotics co-existed as concepts. Several little-known examples significantly predate the 1921 origin of the word robot [Capek], including: the author of Moby Dick [Melville, Herman, The Bell-Tower, 1885], [Adams, Villers de l'Isle, The Future Eve, 1886], [Cook, William Wallace, A Round Trip to the Year 2000, 1903] in which the machines are called 'muglugs", and [Bierce, Ambrose, Maxon's Master, 1909] in which the possibility of a genuine chess-playing automaton was seriously envisioned. It is therefore fitting that AI systems for space, which are robotic in the broad sense that they are part of highly mobile vehicles (spaceships), fulfill the emotional connection of these two concepts in a century of imaginative literature. Science fiction is acknowledged in various ways for having made modern spaceflight possible. For example, there are at least three craters on the moon named for science fiction authors (don't rush to join the list -- you must be deceased to be eligible): Hugo Gernsback (16 August 1884 - 19 August 1967) Willy Ley (2 October 1906 - 24 June 1969) H.G. [Herbert George] Wells (21 September 1866 - 13 August 1946) and at least three craters on Mars are named after science fiction authors: John W. [Wood] Campbell (8 June 1910 - 11 July 1971) Stanley G. Weinbaum (1900 - 14 December 1935), and again H.G. [Herbert George] Wells (21 September 1866 - 13 August 1946) Hugo Gernsback is the man who created modern science fiction (which he called scientifiction) by founding the influential Amazing Stories Magazine in April 1926. Willy Ley was a science writer who came over with Werner Von Braun, and one author of this paper (JVP) met him some 30 years ago and was greatly influenced by his work. John W. Campbell refined modern science fiction by being the single most influential editor in the field, at the helm of Astounding Science Fiction (now Analog) from October 1937 until his death, and who had his first story published while still an undergraduate at MIT [Campbell 1930]. Stanley G. [Grauman] Weinbaum died while just beginning a meteoric career. He was one of the first to describe intelligent beings in the solar system [Weinbaum, Stanley G., "A Martian Odyssey", Wonder Stories, July 1934 (see Martian Odyssey and other Classics of Science Fiction, ed. S. Moskowitz, Lancer, 1962 for more on Weinbaum)] whose intelligence is fundementally different from (even unintelligible to) human beings. This strikes at the heart of the grand dream of AI. Intelligence is not necessarily an imitation of human thought processes. A full examination of the performance and characteristics of AI systems in space -- as portrayed in science fiction -- will be presented (someday, somewhere) in a sequel to this column. Without the history of space, computers, AI, and robots in science fiction presented here, there is no context in which such an examination is meaningful. Since real technology has advanced so rapidly, with space AI systems drawing on several simultaneous conceptual revolutions, it is not surprising that the most sophisticated Space AI fiction is being written today by a dazzling group of newer writers. Teleoperation and telerobotics have themselves become significant themes in science fictionb [Mixon, Laura J., Glass Houses, NY: Tor, May 1992]. Gregory Benford is part way into a trilogy [Benford, Gregory, Tides of Light, NY: Bantam, 1989] set amid the AI civilizations at the heart of the Milky Way galaxy, in which humans are annoying parasites to be exterminated. [Platt, Charles, The Silicon Man, NY: Bantam Spectra, March 1991] examines the implications of downloading an entire human mentalality into a semiconductor substrate. In his acknowledgment, Platt notes "Jonathan Post advised me on the policies and procedures of aerospace contractors" -- so science fiction and space AI really do influence each other. Criminals in the future, rather than being executed, might be downloaded into Space AI systems for exploring and exploiting the solar system [Jennings, Philip C., The Bug Life Chronicles, NY: Baen, 1989]. Downloading might require slicing and destroying the original human brain -- robots might not understand why people resist this approach to immortality [Rucker, Rudy, Software, NY: Ace, 1982]. One of this paper’s authors wrote a science fiction story [Post, Jonathan V., "Brainsails", 1990, in press, and excerpted in the Pasadena Star-News, January 1993] about the combination of space telerobotics and magnetoencephalography. Artificially Intelligent solar sails may travel through and beyond the solar system, as realized in poetry in a book whose profits support a real spaceflight [Post, Jonathan V. and Bradbury, Ray, "To Sail Beyond the Sun: A Luminous Collage", in Project Solar Sail, edited by Arthur C. Clarke, David Brin and Jonathan V. Post, NY: Roc (Penguin USA), April 1991; unabridged version in the 1992 Rhysling Anthology]. Unmanned interstellar probes may require hybrid software which combines conventional AI, plus a simulation of a subjective human personality, plus nanotechnology effectors, to implement what we called a "replibot" [Bear, Greg, Queen of Angels, NY: Tor, 1990]. Space AI may be needed [Sheffield, Charles, Cold as Ice, NY: Tor, June 1992] to process data from 'the Distributed Observation System" -- a network of 500,000 telescopes spread from Mercury to Jupiter orbit. One journalist-turned-SF-author has editorialized (in an expansion of an idea by Marvin Minsky at MIT) that we should abandon our science fictional romantic desires for manned flight altogether, and make a cost-effective investment in planetary telerobotics and space AI interfaced with the public through virtual reality [Sterling, Bruce, “Outer Cyberspace”, Fantasy & Science Fiction, June 1992]. Then, of course, there are the movies and television, which arguably have had the greatest impact on how our culture has perceived the growing use of computers and AI. Several films, such as Silent Running and Star Wars, feature nonhumanoid robots assisting humans in space, a role that much of our culture seems to accept as the most likely future scenario. However, even though our technology will likely make such human-robot scenarios feasible, the need to have humans in space at all will become a growing issue as AI’s power increases. Such an “AI only” view (i.e., using only artificial hardware and software in space) has received increasing attention from SF writers today. Although this scenario would depart from our “humans conquering space by being in space” paradigm, it would likely be safer, less costly, and ultimately more democratic if combined with telepresence’s ability to let everyone share in the exploration . As of May 1992, the NASA channel is already being planned for wider cable access in the U.S.; perhaps this is just the first step. Future couch potatoes might spend a day making new remote space observations on Europa (using AI in the TV to alert them to images that match interesting criteria), flicking the remote (pun intended?) to see how Mars is doing, and still have time left to watch MTV. Two researchers, located on different parts of the globe (or one on Earth, one on the Moon), could use VR gear to virtually explore a more distant orb together, their telepresences being in the same remote locale although they were not. Going further, if these two were a couple, they could switch to another remote planet and have sex, virtually -- without leaving their home or physically touching their partners! (Then they could complain to their artificial shrinks -- “We’re just not close anymore, Eliza”). Note that a slightly different, darker take on this remote exploration via AI was put forth on a recent Star Trek episode, in which an alien artificially increased the intelligence of other races (in this case, a human on the Enterprise) to enable them to make inventive leaps necessary to visit the alien’s world. A revised version of this seems a worthy and feasible goal for our own AI. For instance, we might be able to build robots intelligent enough to utilize elements encountered in space to better their collective situation -- e.g., to replicate themselves, or even to improve themselves (with tasks such as detecting and extracting more efficient fuel, or simply gaining more of their current fuel). Even if we simply send out multitudes of robots and get back lots of visual data, telepresence experiences, and returned samples, we would still be “bringing space to us” without leaving our home, as did the alien in that Trek episode. While many androids have appeared in film and TV, an interesting case is Ulysses from Making Mr. Right; this droid was designed for space travel as a safe substitute for man, superior in mental facility as well as in its immunity to loneliness and other human “frailties”. However, Ulysses, once sentient, soon wants to experience love and stay on Earth. This raises an interesting question: might AI progress to the point where an intelligence might not want to follow its programming (i.e., its orders/mission)? This issue was also raised in Trek, when Data (another android longing for humanity) is ordered to be dismantled for scientific study but a court rules him to be alive, sentient, and worthy of rights. And of course there is HAL in 2001, which decided to deviate from some of its orders as well. In short, there may come a point in the future when software designers will have the twin constraints of needing software that is intelligent enough to do vital tasks, but not smart enough to decide that these tasks are not worth doing! But with the advent of learning algorithms today, and their inevitable improvement, will absolute control of AI programs even be possible in later years? When Stanley G. [Grauman] Weinbaum was one of the first to describe intelligent beings in the solar system [Weinbaum 1934] with intelligence fundamentally different from (even unintelligible to) human beings, thus striking at the heart of the grand dream of AI -- intelligence is not necessarily an imitation of human thought processes. This raises issues such as this: if advanced AI is used on a long space mission, and its knowledge is altered greatly during its duration, we might not be able to comprehend its output after a certain point in time (assuming a suitably vast capacity to learn new concepts). A related theme was the heart of the first Trek film, in which Voyager is redesigned into a super-intelligent entity by an alien being. If we replace the alien by advanced (albeit terrestrial) learning algorithms, the potential for the evolution of a craft’s knowledge beyond our understanding is at least a possibility. Such a result occurred in the film The Forbin Project, based on the Colossus novel; humans stopped eavesdropping on the superintelligent computer once it evolved a language beyond our comprehension. But remember that the need for many space travel “solutions”, such as altering humans, could be eliminated if, as we mentioned near the beginning of this section, we employ remote space exploration using AI. In this regard, an irony exists in one of Trek’s most intriguing “inventions” -- the holodeck, a hypothetical combination of Virtual Reality and AI (a synthesis that one author coined VRAI [Post, 1990], the French word for “truth”). In a holodeck, one can sample alternate worlds (VR) as well as interact with characters in that world (AI) that can be programmed to one’s wishes. Such technology is not that far off from (real) reality today, and its advantages would be numerous for space domains. For instance, an astronaut -- or even an Earth-bound “teleexplorer” -- could try out fixes to spacecraft hardware in a holodeck without having to alter the actual device. In addition, stored technical experts could be “called up” out of digital hibernation and consulted even if they are physically distant (e.g., deceased). Note that an artificial expert could even be a composite of several people’s knowledge and interpersonal styles -- a being who never actually existed in the real world (in that particular “configuration”). It is ironic that the holodeck idea, developed for use on a (fictitious) space vessel, might one day be used to enable humans to avoid actual travel at all (and perhaps may make the notion of large spacecraft obsolete). If this permutation of the initial holodeck intent eventually gets used for teleexploration, it would not be the first time an SF idea led in a nonlinear path to real payoffs. (Besides, who would have watched “Star Trek: The Stay-at-home Generation”?) 8. Conclusions and Final Thoughts Where will the “real” future lead us? Given the above examples, scenarios and discussions, certain trends are more prevalent than others, and we can venture some final predictions, adding to those already presented. The trend in robotics and AI software for space can be summarized with keywords such as small, cheap, flexible, adaptive and autonomous, as well as large numbers, redundancy, decentralized intelligence, remote operation and global teleaccess. Also, as the applications described earlier become increasingly common, powerful, and less expensive, synergy seems inevitable. In both robots and AI software, the use of autonomous independent intelligent agents should enable an increasing number of functions to be performed continuously, with little human intervention. Such agents would be the software analog to the dumbots, in that they would represent specific specialized collections of knowledge and processes that “live” on their own, gather their own input, and communicate (to other software or to humans) when the appropriate conditions arise. Machine learning methods should allow these agents to improve their behavior during long missions. In fact, learning will prove essential to deal with the unknowns of unexplored space, since no mission planner can predict all required system reactions, and instructions from Earth are impractical for long distant missions. In time, fewer astronauts should be required per mission, increasingly replaced not only by telepresence equipment but by "astronauts on a disk". Such "astrobots" would have all the usual stereotypical benefits of automated workers -- more vigilant, more efficient, no sleep requirement, faster, able to free any astronauts for other tasks, etc. The longer the duration of future missions, and the higher the chance of danger, the more valuable these automated astronauts would be to the success of such missions. In summary, astrobots should reduce human risk by decreasing the number of humans required for a given mission, and by allowing those remaining (if any) to perform tasks with greater safety and probability-of-success -- via consultation of automated expert systems, telepresence using the ship as the base (e.g., when Earth links are not feasible), and other techniques. Perhaps Domino, the robot in [Budrys, Algis J., Michaelmas, NY:Berkley, July 1978], puts it in a more imaginative light: "My bones are made of steel The pain I feel is rust. The dust to which your pangs bequeath the rots that flourish underneath the living flesh is not for me. Time's tick is but the breathing of a clock. No brazen shock of expiration tolls for me. Error unsound is my demise. The worm we share is lies." In summary, we presented several examples -- a survey of ideas and technologies -- to illustrate where AI has been applied in the past, some of today's issues, and ideas regarding how it might be applied to space missions of the future. AI should enable an increasing number of future missions to pose reduced risk to human lives, increase the amount of exploration that can be done without leaving Earth, and enhance the effectiveness of missions in which we or our surrogates go where no one has gone before. I don't know where this column will go, now that Quantum lives only in the archives and your memories, but here's hoping that you and I, or at least your AI extensions and my hypertext output, will meet again and again. The universe may or may not be infinite, but let us act in the belief that our imagination has no Outer Limits.