the Rev. Edmund Robinson
Unitarian Universalist Meeting House
May 20, 2018
You cannot get through a day’s news without encountering something about Artificial Intelligence. Just Friday, the New York Times had an article about a visit to MIT by the Canadian Prime Minister Justin Trudeau. Trudeau said Canada had invested massively in developing AI and was prepared to invest more, and that “Canada is making a ‘deliberate choice,’ but not an easy choice, to embrace change at a time when new technology is disrupting workplaces and leading to anxiety and fear about the future.” Saturday in the same paper, an op-ed by a psychologist and a computer scientist assured us that AI is a lot harder than you think and we are a long way from actually creating intelligent beings. Over in the Atlantic magazine, we have a new article by none other than Henry Kissinger titled “How the Enlightenment Ends,” where he argues that “[p]hilosophically, intellectually—in every way—human society is unprepared for the rise of artificial intelligence.”
As most of you know, I’ve been asked to preach to the annual gathering of IRAS, the Institute on Religion in an Age of Science, in late June on Star Island as they deliberate artificial intelligence. Today’s sermon is one of several trial runs for that preaching, but I wanted to tailor it to the Social Justice Committee’s determination that May would be the month of the woman. So I want to talk about a particular feature of Artificial Intelligence that has come to the fore lately: fembots, or artificial intelligence devices – “people,” sort of – which are deigned to appear to be female.
The first recent one was just a voice. In the 2013 film “Her,” a man has a sort of love affair with the new operating system of his computer, which has a female voice. This robot has no body at all, but still manages to sort of seduce the hero.
In the 2014 movie Ex Machina, a young man who is the brightest programmer in a software company is invited to the secluded compound of the very rich C.E.O. of the company for a mission revealed to him only after he gets there. He is to perform a Turing test with the latest robot produced by the CEO, who is in the form of a beautiful woman – that’s her on the Order of Service cover.
A Turing test, a challenge created in 1953 by the early computer genius Alan Turing, says that we will have achieved artificial intelligence when a human, in asking any questions he wants, can’t tell if he is communicating with a machine or another human, or in other words, that the computer is able to simulate a human mind in communication with a real human. In the film Ex Machina, the Turing test amounts to having the hero fall in love with the robot, who deceives him and kills the CEO and entraps the hero in order to make an escape from the compound.
Another deadly fembot is in the HBO series Westworld, which takes place in a Wild West Theme park staffed almost entirely by robots. The robots are specially manufactured to look and act like Westerners of the nineteenth century. The idea of the park is that paying guests in the present time can live out their violent fantasies by apparently killing the robots, but the bots can simply be patched back up and returned to service. But of course the bots go haywire and turn on the humans in a twist beloved of science fiction writers. The principal heroine is a lovely lady with long blond pigtails who is sweet and takes care of her father on the old ranch, but who turns into the principal villainess.
These are all fictional robots. But some actual female robots are being made.
Last week, we talked about wisdom from our mothers, which in Greek is Sophia. That is a very important concept in philosophy (a word which means love of wisdom) and religion. But when you Google the term Sophia now, the first thing that pops up is a fembot. And you met her earlier, on the Jimmy Fallon show.
Now let’s do a Turing test here: how many of you thought Sophia was a real human being? How many of you thought Poppy was ? My take is that Sophia was a robot trying to act as human. And Poppy is a human actress trying to act like a robot. But I could be wrong.
Many bots don’t have a visible existence. Most AI, artificial intelligence, which is being developed now, consists of bits of software you can’t see or touch designed to do a specific thing. This kind of AI is big business, because everyone knows the value of automating tasks that are now done by human beings in real time. The more such tasks you can farm out to machines, using algorithms, the more you can free up human time for more productive labor.
But these common bots are not very exciting. Google trotted out a system the other day which phones a restaurant, talks to the human on the other end and makes a reservations for you. It’s an impressive feat in AI, but no one is going to write a feature film or cable TV show about bots like this, however useful they may become. What is exciting and disturbing are the humanoid robots, and those roboticists interested in making a sensation are turning to human-looking robots and choose female models for them. Why? I think the answer to that question has a lot to do with patriarchy and the fact that, after decades of feminism, it still persists so strongly.
Before we get to the physical package in which the intelligence is set, we need to say a word about the intelligence itself. Artificial Intelligence, or AI, is a broad general term for any means of doing an intellectual task with machines. Google’s search engine, Facebook’s ability to identify people in photographs, language recognition software and language translation, the personal assistant in your phone or in your speaker which can speak to you and act on your response, all are bits of artificial intelligence. Some of them are quite powerful. We have chess programs which can beat human grandmaster players, we even have computers which can beat masters at the game of Go, which is much more dependent on judgment than chess is.
But outside of its area of expertise, each of these bits of software is stupid. Some would say that the holy grail of Artificial Intelligence is what is called AGI, artificial general intelligence. Can we build a machine that would be able to think for itself and act for itself, that would have agency, the capacity for self-reflection, the ability to empathize and make reasonable guesses as to the intentions of other people, who would have an ethic and act with moral responsibility? Could we have a machine that in some sense knows it is alive, which thinks and knows that it is thinking?
We might be tempted to ask the question, can a machine be conscious? However, people in the field say this is the wrong question, because what the attempt to make a machine conscious has taught us is that we are mistaken to believe that humans are conscious. We can’t reliably put our finger on what consciousness is, and until we understand what it is we can’t replicate it in machines.
When I was a youngster I was fascinated with robots. At about age nine, I got a toy one for Christmas. As a child I would have thought of it as genderless. Robots, I would have reasoned, are made to serve humans, to do tasks for humans, and there would be no point in assigning the human physical characteristics of sex or the cultural characteristics we call gender to them. By analogy, at the point when I got that robot toy, I was a believing Episcopalian, and had been taught that God had no sex or gender. The Creator can create through unlimited power, does not need to procreate using the biological equipment of humans and other animals, does not need a body of any kind, much less one which fits into the types we think of as male and female. If the being which created us does not need a gender, then the robots, or beings we create, do not need them either.
That thinking to me now seems rationalistic and sexist. The Bible says God created humans in God’s own image, and whether or not that is so, we have certainly imagined God in human form, and much of the misery of the human race is traceable to male dominance which has been supported theologically by the idea of a male, omnipotent God. So if an AI company wants to package its intelligence apparatus in a human-like form, there is about a 50-50 chance, statistically, that they would choose a female one. My reaction of surprise to seeing a female robot is therefore sexist. If we are setting about to replicate humans, half of them should be female.
And yet there is another level to this. This happens at a time, particularly in the upcoming generations, where a bunch of people are questioning the traditional division of the human race into men and women. Many refuse to go along with assignment to one gender or the other, insist on using neuter or plural pronouns to describe themselves, choose to live with gender fluidity. Us older folks were brought up to assume a binary human race, and the younger folks say, “why just two?”
Why just two, indeed? Human biology may place limits on the configurations of our bodies, but those limits wouldn’t apply to robots. A robot manufacturer, designing a human-like package for the artificial intelligence system it has created, is free to design any form it likes.
However, a robotics company’s ability to sell robots is limited by human acceptance. Trans humans who present with visible characteristics of both traditional sexes often find it hard to win acceptance in today’s society, and there would be little incentive for a company to manufacture robots which did the same.
In short, the whole notion of gender is in a state of flux in society at large, and artificial intelligence comes on the scene into this ambiguity and raises significant questions.
We have focused on the physical packaging to this point, but what about the artificial intelligence itself? Is there a specifically women’s intelligence and men’s intelligence? When we pose a question like that, we have to immediately understand that there are a bunch of assumptions and stereotypes out in the culture which may or may not be supported by rigorous psychological testing. Men are from Mars, women are from Venus. Women are instinctively in touch with their feelings, men are alienated from them. Or as it had been expressed, men talk about baseball to avoid talking about their feelings, while women talk about their feelings to avoid talking about baseball.
A recent BBC piece summarized a decade of massive psychological studies across many different cultures to conclude that men and women indeed do have different personalities. Women are generally warmer, friendlier, more anxious and more sensitive to their feelings than men, while men were more assertive and open to new ideas.
Yet the deadly fembots of Her, Ex Machina, and Westworld defy this stereotype; they turn from empathetic, warm and even cuddly creations to unemotional, cold killers. And this defiance of our expectations is what gives them the deadly fembots shock value. And shock value of any film pays off at the box office.
As I noted in a sermon three weeks ago, the idea of a robot uprising where fully conscious artificial beings turn on their masters did not start in the twenty-first century.
Our word “robot” comes from a Slavic word meaning slave, but it was first used in a Czech play in 1920 by Karel Capek called R.U.R. The title stands for Rossum’s Universal Robots, and the play is set in a future where robots are manufactured to serve in ordinary jobs. They are made in a factory, not with metal parts or silicon chips but somehow with flesh and blood. The basic ethical questions posed by such manufactured beings are all in this early play:
(1) do they have consciousness?
(2) if so, is it murder to turn one off?
(3) is it murder to reprogram or otherwise recycle them when their usefulness in a particular role has been compromised?
(4) is it slavery to set them to work without any personal compensation and work them around the clock just because they don’t need food, sleep or shelter?
(5) if they are sentient beings but are not humans, do they have human rights?
And the robots in Rossum’s company end up revolting and killing their human masters. So the idea of the robot revolt is not new. Nor is the idea of female robots, for the first robot we meet in the play R.U.R is the secretary in the front office.
A few weeks ago we passed the fiftieth anniversary of the release of 2001, a Space Odyssey. The computer villain in that movie was named Hal, and though Hal did not have any kind of body – all we saw on screen was a big red eye – we can assume from the voice and name that Hal was intended to be viewed as male. Hal is terrifying because he speaks in a calm, measured voice as he goes about methodically killing the human inhabitants of the spacecraft until the hero manages to disconnect his higher intellectual functions. Hal is in many ways the picture of reason in 1968, the year the film came out; reason disconnected from emotion.
We know a lot more about neuroscience now than we did in 1968, and one of the things we have discovered is that Descartes was wrong when he said, I think, therefore I am. Thinking cannot, in the real brain, be separated from feeling. It is more accurate to say, I feel, therefore I am, or I fear, therefore I am, I love, therefore I am, I enjoy, therefore I am.
Feeling, in turn, in the human as well as in all animal species we study, is intimately connected to states of the body. If I am sad, happy, elated, wistful, depressed, angry – whatever the feeling I am experiencing, it is as likely to be connected to what’s happening with my body as to what’s happening in my life.
We know that the brain is not just a computer. In fact the very word computer is kind of misleading for we now do much more with these machines than number crunching. In some ways, as I have said, they have surpassed human intelligence.
But human intelligence is inextricably linked to human emotion and human emotion is inextricably linked to the body. If we want to make artificial replicas of human beings, we are going to have to give them human, biological bodies or come up with some substitute.
But here is the rub, and the point of these reflections: maybe the whole AGI enterprise, the whole effort to create a stand-alone artificial consciousness is an attempt to transcend emotion altogether, to take intelligence into a stratosphere where it will not have to deal with petty distractions like sadness or elation or depression. In this view, unemotional intelligence is superior intelligence.
But this view is wrong from a neurological perspective; brains do not work in a vat unconnected to a body. You won’t be able to cheat death by uploading your consciousness to a super computer any time in the future. And the vision that you can, or it would be desirable if you could, strikes me as a male vision.
I am a male, and to some extent find myself trapped in a male outlook. Part of me says the plotlines which have us conduct the Turing test by falling in love with machines are absurd and ridiculous, and we deserve to be shocked at the end when the fembot turns out to be heartless. We have not fallen in love with a person, we have fallen in love with falling in love. The only being on the other end of our emotion is about as human as a washing machine. We can have strong emotions about inanimate objects; some of us feel warmly about our cars, our cellphones and laptops, our houses, but we don’t mistake them for people.
Yet part of me applauds the very attempt to put some emotion into AI. Neuroscience, what we know about how actual minds work, shows that any attempt to artificially create a human mind without a human body – the whole wet gooey package with its peculiar chemistry, not just the outer form – is doomed. And why would anyone want to create another being who has to be fed and go to the bathroom and sleep? Isn’t the whole point of artificial beings that they transcend human limitations?
In the end this Hollywood creation of fembots may bring us to appreciate what it is which makes us all human.