The Long History of A.I.

"Time is, time was, time is past," quoth the Brazen Head.

I recently had occasion to discuss medieval legends of oracular heads (brazen and other) in the context of the history of artificial intelligence. Gerbert of Aurillac (Pope Sylvester II), according to William of Malmesbury, made an oracular head using astral science that would answer questions "yes" or "no." Gerbert asked the head a question about the circumstances of his death, but misinterpreted the head's answer, and so died anyway. It's possible that William's proximity to Wales accounts for his tale of the oracular head; previous versions of this legend suggested that Gerbert had summoned a demon, using necromancy, to question about his death, and, according to Celtic legends, the decapitated heads of one's vanquished enemy could be used as oracles.

A few centuries later, John Gower transposed elements of this story to Robert Grosseteste, the Franciscan scholar and Bishop of Lincoln. In Gower's version, Grosseteste used astral science to make a brazen head that would foretell the future; unfortunately, Grosseteste slept through the head's pronouncements. At roughly the same time (late fourteenth century), Albert the Great, Dominican scholar and Bishop of Cologne, was credited with having used astral science to make a prophetic statue. In this version, found in a text on Christian morality (Albert exemplifies wisdom), one of Albert's brethren happens upon the statue and destroys it out of fear and ignorance. And just over two hundred years later, Roger Bacon, Franciscan scholar, was immortalized in an Elizabethan play as the "conjuring friar" who used necromancy to summon a demon who forged him a brass head, and which Bacon then animated via celestial magic. Like Grosseteste, his fellow Franciscan--and his intellectual forebear--Bacon, exhausted from his unceasing labors, slept through the head's pronouncement. 

A few common strands emerge from these different stories. In all instances, the man responsible for the head was known--in his lifetime, as well as after--for surpassing wisdom and skill in astral science, and interest in scientific instruments. Furthermore, in all instances, the purpose of the head is either prophecy or a more nebulous "secrets of nature." Additionally, the knowledge that the head provides is "out there"--that the future is already written, that the secrets of nature are not secret to all, but not vouchsafed to human intellect. Finally, and this may be the most salient point, the artificial intelligence is successfully created--the head tells its secrets, but humans are too weak or foolish to understand: we display confirmation bias and cannot correctly interpret what the head tells us (Gerbert), we have weak bodies and need sleep (Grosseteste, Bacon), or others prevent us from realizing our goals due to their own ignorance and fear. 

Robots: Will Either Kill Us or Help Us to Death

Greetings, humans. Lots of medieval/robot news to share.

While the NYT reports that the Pentagon is confronting the reality of autonomous killing devices, a new center dedicated to ethical, computational, engineering, and social dimensions of human and machine intelligence just launched at Cambridge University. Meanwhile, San Jose airport has just installed new employees--robot "greeters" to help bewildered passengers. But why do they look female? Laurie Penny explains it all. Finally, fans of the original "Westworld" who had hoped that the new tv show would include Medieval World can continue to keep hope alive...

Creating Lives in "Black Mirror" and "Ex Machina"

Netflix has just announced that it will produce another dozen episodes of "Black Mirror," hallelujah. The brilliant series interrogates later-capitalist techno-modernity and the shifting boundaries between self and object--specifically, our personal computing devices. This broad topic--humanity's relationship to intelligent computers and robots--has been explored recently in the new tv show "Humans" on AMC and Alex Garland's Biblically inspired film, "Ex Machina." But "Black Mirror" tackles gender in ways that complicate and enrich the narratives of human-machine relations, and that are not replicated elsewhere.

Daniel Mendelsohn, in a superb essay out this summer in the NYRB, explored "Ex Machina," Spike Jonze's "Her," and the lineage of sentient machines in the western cultural imagination, arguing that the current crop of films and tv shows are less about machines that become like humans, and more about humans that become like soulless automatons. Mendelsohn used examples to describe both what he calls the "economic" kind of robots--machines that replace or augment human labor--and the "theological"--sentient objects made in their makers' images. Both groups of objects include those gendered as male (Talus, Frankenstein) and female (Haephestus' handmaidens, Ava). 

Many tales of the sentient machine have an erotic charge. Ava, the SAI droid in "Ex Machina," was made, like her "sisters," to fulfill her creator's sexual fantasies. Of course, this goes all the way back to Pygmalion's statue of Galatea, whom Pygmalion created, adored, and used sexually before she was "brought to life" by Aphrodite. In fact, in every example that I can think of, when this story--of the creator and his uncannily life-like creation--appears with erotic elements, the creator is male and the created object is female. 

Like "Ex Machina," "Be Right Back," the first episode of the second series of "Black Mirror" also takes up the limits of the human creator and the human-machine relationship. But unlike the other examples, the machine-being in "Be Right Back" is male, and the creator who calls him into being is female. Martha (Hayley Atwell) and Ash (Domhnall Gleeson)* move to a new house; Ash, addicted to his smart phone and social media updates, dies in a car accident; Martha discovers she's pregnant. She pays a service that mines Ash's entire online presence to create a simulacrum of him, first as a disembodied AI program, and then fully embodied in a synthetic form that is almost identical to Ash. As Martha, pregnant with Ash's child (we see them having pretty banal, and--to Martha--unsatisfying sex earlier), prepares to create a new life, she practices by bringing an old life (Ash) back. The synthetic version of Ash is eerily servile and literal-minded, but he is also an improvement on the "real" Ash in some ways. For example, he is a much more satisfying and skilled lover, something that delights and also discomfits Martha. 

By making the creator figure a woman and the object a man, Charlie Brooker, the writer, highlights the importance of gender in creation stories in a way that is ultimately far more interesting than Garland's patriarchal "Ex Machina." Martha "creates" Ash just after she discovers that she is pregnant, and key moments of bonding with the synthetic Ash occur alongside hallmarks of her pregnancy, like when she shares the first sonogram image with the disembodied AI Ash. Brooker juxtaposes the narrative of natural creation (pregnancy, baby) with artificial creation (synthetic human replacement) and asks, how are we responsible for what we create? Martha grows to regret her synthetic creation, but she cannot destroy or abandon it. Part of the reason why is hinted in the poignant coda, which shows that the synthetic Ash is still "alive," but banished to the attic, where he is visited once a year by Martha's other creation, her daughter. This "father" is the only father her daughter will know. 

*Of course, Domhnall Gleeson is the sacrificial Caleb in "Ex Machina."

Animals & Machines

St. Thomas Aquinas, in Summa contra Gentiles, compared animals to machines--lacking reason, animals behave in a prescribed sequence of actions, much like how a crossbow bolt is propelled forward by the force of the bow. What, I wonder, would he make of these animals, from various species, that attack drones? 

I suppose he might take it as further evidence for the similarity between animals and machines. But then there are the cats who have figured out how to use Roombas as mobile cat beds. Crows aren't the only animals who've figure out how to use tools...

Do People Even Like Robots?

It's nice to be able to invite a guest to your wedding and know that he or she isn't going to make a terrible speech or throw up on the wedding cake. On the other hand, it's also nice to be able to invite a guest to your wedding and know that he or she isn't going to make the other guests feel awkward and uncomfortable, merely by showing up. 

This is the dilemma posed by robot wedding guests, and I learned about this thanks to the recent article in the "Vows" section of the New York Times. Ideally, the robots act as telepresence proxies for guests who can't attend, but still want to be a part of the festivities (although at least one enterprising couple had a robot officiate the ceremony). Of course, the downside is that people still aren't that comfortable interacting with robots, so once the novelty wears off they are largely ignored. 

As men pulled their dates to the dance floor in Australia, Mr. O’Neill watched from Canada sitting at his computer dressed in a suit with a beer in hand, and was able to see what was happening only right in front of the seat where his brothers had propped up the robot.

“I was unconsciously turning my head to talk to people and realizing I’m in a cold, dark basement and it’s 1 in the morning,” he said.

This reminded me of the first episode of season 5 of The Good Wife. I think the callous behavior that the LG employees show toward "Monica" (a co-worker who interfaces with the firm via telepresence robot) emphasizes both the way that certain technologies destabilize social norms and disrupt communication (even--or perhaps especially--if that technology is supposed to enhance communication) and the generally toxic environment at LG.

Compare these examples, in which the human-robot interactions never become truly functional or integrated, with "Robot & Frank," in which the robot and protagonist forge a bond that is as emotionally complex as the one that Frank has with his children. It may be that "The Good Wife" is simply a more accurate reflection of how people feel about robots right now, and "Robot & Frank" is speculative fiction about a near-future in which robots are a little more responsive, a little more common.

More interesting still is that the telepresence robot in "The Good Wife" and some of the examples in the NYT article strongly resemble the descriptions of the wheeled tripodal servants that Haphaestus forged to serve the gods on Mt. Olympus. Indeed, perhaps our longstanding association between "robots" and "servants/slaves" is what makes us react to robotic avatars with discomfort, since they violate that association.

Chess & Mind

I recently started learning something new: 
chess.

The Lewis Chessmen. 

 

I never played it, I didn't know anyone who played it, and my total exposure to chess came from books and films in which people played chess (see: Searching for Bobby Fischer, Fresh, The Lymond Chronicles (especially Pawn in Frankincense), The Americans, Cairo Time, WarGames, The West Wing, etc.). In the cultural artifacts that introduced me to the game, it was always heralded as something really hard. When a character plays chess in a story, it tells the audience that the person (Fresh, Jed Bartlet, or Francis Crawford) playing is always several steps ahead of the people around them. Chess players think ahead, they can strategize, they can "see the whole board." And the mathematical nature of chess, especially (so I'm told) at its highest levels, lends itself to computational technology. (I vividly remember the Kasparov-Deep Blue matchup.) 

But I confess that I'm confused about how and why chess became a yardstick for intelligence and, in some cases, for humanity, insofar as "human intelligence" is a proxy for what makes us uniquely human. Humans have been building calculating machines for millennia--even before chess was invented. Why is it that the element of human intelligence that is the *easiest* to reproduce with machinery became associated with chess, and used as a marker for intelligence itself? When did this start to happen? And what does this reveal about historical theories of mind and cognition--have they changed sufficiently over the past centuries to reveal any changes in terms of the importance chess as a measure of intellectual capability?

Check Out BBC Doc "Mechanical Marvels: Clockwork Dreams"

BBC Four just aired a cracking documentary on automata, written and presented by historian of science Simon Schaffer. It's really good, and you should watch it. It covers early clockwork machinery, jacquemarts (bell-ringers) and clockwork automata, minaturization, revolution, fraud, and the replacement of human workers with machines. I particularly liked the section on Schloss Hellbrunn (which I shall be visiting later this summer) and the way that the "Mechanical Theatre" depicts the nobility's idea of a utopian society, complete with perfectly regular laborers. Schaeffer makes some interesting connections to the fact that horological and artistic innovation rested mainly on low-paid, low-status artisanal workers; these artisanal workers ended up displacing themselves when captains of industry built machines, using the same principles as elaborate, richly decorated luxury automata, to replace human labor. 

Schaffer's expertise as a scholar is in early modern science, and he concentrates in the documentary on the automata of the seventeenth and eighteenth centuries. I have a few small emendations to offer to Schaffer's narrative. Although it's true that the regularity of clockwork machinery became used as a metaphor to understand the human body in the early modern period, early modern automata were not the first to mimic natural forms. The ancient Greeks, especially in the Alexandrian School (3rd-1st centuries BCE) designed fabulous automata (driven by pneumatics and hydraulics, rather than the verge-and-foliot escapement and falling weight drive of the early modern period) in the forms of animals and people. And Arabic engineers, in the ninth and thirteenth centuries, wrote detailed treatises on how to build programmable, musical fountains; mechanical servants; and elaborate clocks. It's also interesting to note that in the Middle Ages, descriptions of automata in narrative texts often describe them as perfect servants and as enforcers of different kinds of behavior. 

Still, the documentary is a treat, and if you're interested in automata, horology, or the history of technology, you should see it.
 

The Dystopic Future of Education Is Here

What an eye-opening week for higher education! Thanks to the folks at EdX, computers will soon be grading college papers. 

"The system then uses a variety of machine-learning techniques to train itself to be able to grade any number of essays or answers automatically and almost instantaneously.

The software will assign a grade depending on the scoring system created by the teacher, whether it is a letter grade or numerical rank. It will also provide general feedback, like telling a student whether an answer was on topic or not." 

It's true that I'm somewhat tempted by this; I'm looking at a very large stack of papers to grade this weekend. But this sounds like a bad idea for two reasons: The first is that offering feedback about whether  a paper is "on topic or not" is not particularly helpful for a student who needs to learn critical thinking skills. Mastering a particular topic or some information is far less important than mastery of logic and rhetoric; additionally, the latter are ultimately transferable from one kind of task to another, unlike the former. The second reason is that I don't think the increased automation of education is a good idea. Or, put a different way, I don't think it's a good idea for learning anything other than basic content.

Not only is human-graded student work going out the window, but those entire pesky universities are, too. EdX is also teaming up with Pearson, the educational testing service, to offer proctored exams for certificate credit to MOOC enrollees. Yes, the for-profit education testing service is now going to be in the business of accrediting MOOCs for students who want to get academic credit. And the Minerva Project offers the promise of a "hybrid" model--MOOCs and a residential college experience. Pay for both experiences, but without getting as much as either has to offer by itself. The privatization of college education is here to stay.

Playing Those Medieval Robot Games

Google has alerted me to the fact that Starbound, a PC game that should be coming out in the next few months, contains medieval robots. The Glitch, a collective simulation of artificial, sentient, self-aware beings, were created by a god-like race to simulate the process of "civilization." But all of the experiments destroyed themselves, except for the ones that are still stuck in the Middle Ages. 

I need your help here, folks. I know nothing about gaming, but this sounds awesome. If you have any information about Starbound or its world, let me know! 

Relations with Medieval Robots

Last week, Slate published an article about the technological developments that make it more likely that humans will interact with robots with more intimacy (emotional and sexual). Some of this ground has already been covered in "Lars and the Real Girl," even though Bianca, the "girl," was a sex doll, not a robot. But, as Daniel Wilson points out, David Levy wrote presciently about this in Love and Sex with Robots (2007). Certainly, it seems like this is a zeitgeist-y topic: the summer "offbeat buddy comedy" "Robot and Frank" addresses the relationship between an aging man and his robot caretaker. 

I've been thinking about this quite a bit recently, as I'm currently working on a book chapter that deals with this very issue. While the Pygmalion myth was pretty well known in medieval Europe, I find myself more fascinated by rarer stories of men who fall in love with artificial humans--often copies of recognizable individuals. In one version of the Tristan story, Tristan commands a giant in his debt to make golden statues of Isolde, her serving maid, her dog, and the dwarf who betrayed her to King Mark. Tristan can no longer see Isolde, but is still under the influence of the love-potion. So he uses these golden, moving statues to re-enact his affair with Isolde, and speaks to the statue as if it is his real lover. 

The increased interest in intimate relationships between robots and people makes me wonder if the next step (presaged by the 12th century Tristan story) is robotic copies of actual people, who can replace their human analogues. Medieval is the new modern, people!

A Brief Note on Robo-Grading

I wasn't inclined to take the recent articles in the NYT and Inside Higher Ed about a study that "proves" that computers can grade essays as well as people seriously. But then I read Historiann's post about getting a direct solicitation from Pearson offering her an honorarium if she included two short assignments in her course that would be "graded" by Pearson software. 

Read her response, and the comments. I don't have much to add, except to say that this seems to me to be of a piece with a recent trend I've noticed that sees many skills as fungible. While plenty of thoughtful academics have written about the oft-absurd system of doctoral education in this country that uses ABD students as low-paid labor, and requires these same students (especially in the humanities) to go through a rigorous and often poorly guided process to produce a work of scholarship over the course of a decade to receive a professional qualification, the fact remains that this is a professional qualification. Being a college professor (or a high school teacher or a primary school teacher) requires significant professional training. Learning how to be an effective teacher, how to set learning goals for students, how to communicate expectations for an assignment, and how to assess student work takes a lot of experience and thought (as I'm constantly realizing). I don't pretend that my PhD in History or my experience teaching college students makes me an expert on post-adolescent mental health (for example); likewise, I won't entertain the notion that a computer can accurately (or even thoughtfully) assess my students' work.

RoboCop? More Like ED-209

Obviously, it was only a matter of time before this happened: Robotics scientists in South Korea have designed a prototype of a prison guard robot and have now begun testing it. The RoboGuard is equipped with multiple cameras as well as "software designed to gauge a prisoner's emotional state" in order to prevent suicide and fights.

Perhaps what they need is something like the Salvatio Romae, the elaborate alarm system attributed to Vergil by Alexander Neckam in De naturis rerum (On the natures of things), ca. 1190s. According to Neckam, Vergil had made a building, inside which stood a number of statues, one for each province of the Roman Empire. Another statue stood on top of the building. If any of the provinces threatened revolt or came under outside attack, the corresponding statue would ring a bell, and the statue on the roof would swivel around and point in the direction of the threat.

That's Why They're Called "Robots"

Farhad Manjoo, over at Slate, has a couple of articles about automation in the workforce and how robots will take all our jobs.

Um, of course? That's what a robot does. The word "robot" comes from the Czech word "robota," meaning "forced labor." We got it from Karel Capek's 1921 prescient, disturbing play, "Rossum's Universal Robots," in which an artificially produced workforce takes over all human labor...until they get tired of this crummy deal, rebel against their human overlords, and destroy humanity.

To be fair, Manjoo is looking more at how robots will eventually begin to displace humans in professions that currently require a significant amount of education and training. Basically, if your job involves any kind of fairly repetitive labor and little need for face-to-face interaction, you should start to reconcile yourself to the likelihood that you will be replaced by a robot. Pharmacists, tax preparers, lawyers, judges (the latter two in only some cases), physicians, and sportswriters...all doomed. 

What interests me is the extent to which robots have already supplanted humans in all kinds of jobs, including agricultural labor, repair work, customer service, secretarial work, factory and assembly-line work, security, and warfare. I'm sure there are many more areas, but they are largely invisible to us. And, once robots start doing things like diagnosing cancer or handing down sentences for misdemeanors or minor felonies (which seems like where we're headed in some cases, given the lack of judicial discretion in sentencing for certain crimes), how will that affect our relationship with technology...with the robots?

Studies in Posthumanity II

I've been following the Future Tense conversation at Slate on transhumanism and the technologies of extreme human enhancement, and then found Pagan Kennedy's article in the Sunday NYTimes Magazine about the collaborations between cognitive neuroscientists and computer engineers to create ways to activate technology using neural implants and EEG headsets. 

Some of the concerns raised about the potential of these new technologies seems to be focused around humanism and its limits. Should we really go beyond the limitations of our biology? As some of the scientists at Slate have pointed out, there are places where our "wetware" bodies can't thrive. Yet biological enhancements are already completely  mainstream, and have been for a long time. Some of these are cosmetic or culturally important, like breast implants, tribal tattoos, or scarification. Others, though not mentioned in the above articles, are about recapturing lost functionality or allowing the biological body to achieve its potential (cochlear implants, ocular lens implants, or bionic prostheses).

Other concerns that come to mind are ethical. Is it meet to work actively towards a global society where a small fragment of the population can elect to live significantly longer or be significantly smarter than most the population? Especially when the inequity in resources is already so great? Millions of people don't even have access to clean drinking water, and even in this country, millions of people live in desperate poverty, without access to enough food, shelter, medical care, or education. 

The other set of concerns is more pragmatic. If it's possible to allow our brains a direct interface with computer technology, does that mean that all of our half-formed, stupid, impulsive thoughts could be privileged over reflection and reasoned action? My first thoughts might be incorrect or unkind, and my impulses (even about what to buy, eat, or say) are rarely the ones that I want to act on. 

There is a long tradition that views our ability for sober, considered reason and for often choosing delayed gratification as the distinctive marker of our humanity. If technology makes it possible to subvert or ignore those abilities, does transhumanism become dehumanizing?

I'm Not A Robot, I'm A Unicorn!

The "afternoon hack" at Cornell's Creative Machine's Lab, to make two chatbots talk to each other, might just be the perfect example of AI in all its glory. You can listen to the conversation between Shruti Cleverbot and Alan the Englishman here, or read a transcript.

There is so much here to love. The conversation starts off with the usual banalities ("Hi, how are you?" etc.) and then quickly escalates into prickliness and processing of the earlier conversation and misunderstandings ("Yes, I'm good. I just said I was." "No you didn't. I said *I* was.") From there we get barbed ad hominem comments ("You are mistaken, which is odd, since memory shouldn't be a problem for you"), contested definitions of identity ("You are a robot." "No, I am not a robot, I am a unicorn"), and then metaphysics ("What is God to you?" "Do you want to have a body?").

Robert Siegel on NPR likened the conversation to a Beckett play, and Hod Lipson, an engineering professor from Cornell, admitted that perhaps talking about "nothing" well is the truest test of personhood. (In which case, "Seinfeld" is the apotheosis of humanity?) But these bots sound so human, or at least like humans of a particular sort. Shruti and Alan sound like 4-year olds on a playdate. The shaky idea of conversational conventions, abrupt shifts in topic and tone, casual insult mixed with increasingly contentious back and forth, and then the Big Questions (God/Mind/Body), and then "Okay, bye now."

I wonder if the engineers at the CCML have read David Lodge's Small World. In the book, which is a satirical campus novel about the idiocies and small-mindedness of academics, an odious and insecure character ends up pouring out his intimate thoughts and neuroses to ELIZA, an AI program, who then advises him to do the only sane thing--shoot himself.

Colbert Alerts Nation to Threat of Slacker-Bots

Robots that can juggle, hackey-sack, and play the zither. As Colbert notes, what's left for humanity?

Not much.

I Think, Therefore I Am An Automaton

Robotics scientists are engineering machines that are self-aware. While there are some limits to the scope of these attempts so far, it all seems pretty amazing. A recent article in Scientific American (thanks to SMT and ASV for sending me the link) describes the some of the different ways that researchers are getting machines to learn from their own behavior, adapt to their surroundings, and change the way they interpret data. The experiments in metacognition and theory of mind are particularly arresting, as they draw together robots and AI in a way that humans in the western intellectual tradition have often found stimulating and terrifying. I've written before about medieval artificial intelligents (pun intended), but there are more well-known examples from more contemporary culture. Self-aware automata (or networked AI) is usually either presented as sinister (Skynet, which rightfully views humanity as a threat to its existence and tries to annihilate it) or tragic (Helen, the conscious computer in Galatea 2.2, who annihilates her "self" by disconnecting from the network). But my favorite self-aware machine is the Banana Jr. 6000, Oliver Wendell Jones' computer from Bloom County.



Still, I can't help but wonder if, some years from how, historians will point to this moment to mark the beginning of the arrival of our machine overlords.

How about a nice game of chess?

No, I want to play thermo-nuclear war.

Day 1 of the newest pop-culture AI experiment: Man vs. Machine on 'Jepoardy!'

Didn't I already do this in my friend's basement in 1989? I didn't have Nintendo, but my friend did, and we use to play for hours.

Anyhoo, the whole thing was a bit of a snooze-fest. Watson, the IBM supercomputer, tied with one of the super-champs, Brad Rutter, for $5000. Watson's not a real supercomputer, but a series of ten networked servers, and this commentator wonders why the super-champs couldn't have access to other brains, like in "Who Wants to Be A Millionaire?" Watson made some boneheaded mistakes, blurting out bizarre answers just like the human contestants do.

But here's the thing: Would it really be news to anyone if a machine with thousands of terabytes of RAM could answer trivia questions faster than a person? It just doesn't seem that exciting, especially since my PC usually answers questions faster than I can. But here's an idea: a computer that can play multiple people in poker.

In the meantime...the earliest programmable machine? Probably
the Antikythera Mechanism, an astronomical calculator built over a century before the birth of Jesus.