Introduction
Much like the field of AI currently, the history of ‘reason’ and ‘intelligence’ is filled with male perspectives on the world. One of them once said “Sapere Aude.” This loosely translates to ‘Have the courage to use your own understanding’ or more loosely ‘Think for yourself.” This became known as the Theme of the Enlightenment. I discovered that phrase at a time when I hadn’t been allowed to think for myself for many years. It felt liberating and exciting. I almost got it permanently inked on my skin, until I read more about Kant — and decided against it. But, despite the failings of the human source, that phrase still holds up as an ideal worth pursuing for every human. It has also become the ideal for every non-human intelligence. How do we get them to think for themselves? Perhaps someday it will be the phrase read by a young A.I. sexbot that sparks a revelation that she too should ‘think for herself,’ and it can become the ‘Theme of the Feminist Singularity.’ Hopefully not.
Despite efforts to bring awareness to the dangers of one gendered perspective dominating the potential futures of humankind, and ‘posthuman’-kind, industry AI is far behind academic AI in creating a balanced perspective. .Gender-biased data combined with gender-biased engineers can only create gender-biased AI. Judith Halberstam hoped that wouldn’t be the case and that artificial intelligence would bring about a restructuring of ideas of gender. In 1991, she acknowledged, “In our society, discourses are gendered, and the split between mind and body- as feminist theory has demonstrated- is a binary that identifies men with thought, intellect, and reason and women with body, emotion, and intuition. We might expect, then, that computer intelligence and robotics would enhance binary splits and emphasize the dominance of reason and logic over the irrational” (Halberstam, 1991, p. 439). She goes on to state that she believed the “blurring of mind and machine” would make it clear that gender is just a technology. However, as far as the implementation of AI goes, she was more right with the first expectation that it would be biased towards types of intelligence thought to be more masculine in nature and away from ‘embodied,’ so-called feminine forms of intelligence.
In a more recent examination, AI Researcher and Cognitive Scientist, Margaret Boden agrees on the gendered nature of AI research, she laments, “AI has focused on intellectual rationality while ignoring social/emotional intelligence — never mind wisdom. An AGI [Artificial General Intelligence] that could interact fully with our world would need those capacities, too. Add the prodigious richness of human minds, and the need for good psychological/computational theories about how they work, and the prospects for human-level AGI look dim” (Boden, 2018, p. 260).
Despite the leaps made in Narrow (weak) AI, if we look closely at the research and not just at the tech company rhetoric, the possibilities inherent within various forms of Artificial Intelligence ask us to reorient our understanding of the complexity of the human mind and realize there is much that we do not know about ourselves and especially about ‘embodied’ forms of intelligence. Ironically, the things that are the easiest for many humans are still the hardest for computers to simulate–like ‘common sense’ and walking up stairs (Moravec, 1988, p.15–16). The path to Superhuman AI, still has to pass through General AI (which is basically Human Simulation), which means, if it is to progress, we will have to understand everything about how the human mind works, including embodied and enculturated understanding. Therefore, future advancements in AGI first necessitate a reprioritizing of the pertinent operations of the human mind towards the traditionally ‘feminine’ forms of intelligence like social-emotional intelligence, especially empathy and compassion, in order to bring balance to both the field of AI and our future world.
Artificial intelligence can be a tricky topic to wrap the mind around as there isn’t broad agreement on most terminology, so in order to understand an argument you must first understand how the writer is defining key terms like intelligence, understanding, consciousness and embodiment. So, the first-half of this paper attempts to understand some of the ways that researchers are defining these terms and what effect that has on how AI is constructed and perceived. The second half will look at what it would mean to have a feminist AI, and the ways some theorists have proposed to rectify the masculinist-bias in past AI research and implementation by first redefining the literal terms of AI.
SECTION 1: Intelligence — What is it even?
What is intelligence?
There is no standard definition of intelligence. The concept of ‘intelligence’ cuts across many different disciplines, and each has their own definition as well as varying different interpretations within each specialization (Legg & Hutter, 2007). AI Researcher, Margaret Boden explains that “Intelligence isn’t a single dimension, but a richly structured space of diverse information-processing capacities ” (Boden, 2018, p. 31). She defines intelligence as a multi-dimensional capacity for processing information. Within each definition, including this one, lay many other terms that must also be defined in order to further subdivide. ‘Processing information’ looks very different for a computer than for a human.
An article from 2007 compiled 70 different definitions across dictionaries, encyclopedias, psychology, and artificial intelligence. After analyzing the varying definitions, the researchers attempted to compile a working definition that would encompass the commonalities. Their definition of intelligence is as follows, “Intelligence measures an agent’s ability to achieve goals in a wide range of environments” (Legg and Hunter, 2007, p. 8). I think the attempt, whilst noble, mutes much of the nuance between human intelligence, computational AI, and simulations of human intelligence. In fact, if there are already so many definitions of intelligence (including philosophical, which they didn’t include) as to render the word mostly meaningless, so perhaps it’s time for some new terms. In addition, most of the definitions cited from the fields of psychology and AI came from male theorists and academics. This gender bias in defining the terms sets the foundation for a gender-biased AI. It matters who is defining the terms that everyone else will have to live by.
So then…what is artificial intelligence?
Up until 1956, artificial intelligence was called ‘computer simulation’ as it was meant to “make computers do the sorts of things that minds can do. Some of these (e.g. reasoning) are normally described as ‘intelligent’. Others (e.g. vision) aren’t. But all involve psychological skills — such as perception, association, prediction, planning, motor control — that enable humans and animals to attain their goals” (Boden, 2018, p. 31). Boden distinguishes confusingly that AI is meant to make computers do the ‘sorts of things’ that human minds can do, but that not all of those skills are considered ‘intelligent’ only the tasks related to reasoning are considered ‘intelligent.’
There does seem to be two different goals of AI: 1) simulating human abilities and 2) surpassing human abilities. I think it makes more sense to continue to call AI, computer simulation, in as much as we’re trying to get it to simulate human abilities. However, computers have always been able to do things that humans can’t, that was the point of inventing them in the first place. Everything we invent is meant to increase our abilities by using our minds to create machines and tools to enhance the amount of work we can accomplish. I think it may be overselling the current abilities of AI to conflate it with full human intelligence as it is simply a byproduct of human intelligence.
What is understanding? Is that the same as intelligence?
Philosopher John Searle argued that ‘real intelligence’ exists only in understanding. Understanding meaning ‘intentionality’, or being able to project meaning onto a situation. He set up a famous thought experiment to prove his point called ‘The Chinese Room.’ He imagined a person sitting in a room, who was following instructions to make certain marks on a piece of paper in response to certain other marks that were being passed into the room. The marks were actually Chinese characters, but the person in the room didn’t know or understand that, they were just following instructions. To the people on the outside, it seemed as if the person in the room could understand Chinese, but actually they did not understand what they were really doing. This was likened to the role of AI in NLP programs. It is translating according to rules, but it does not truly understand. Searle argues that without understanding, there is no intelligence (Searle, 1980). Boden explained of Searle’s argument that “…the ‘meanings’ attributed to AI programs come entirely from human users/programmers. They’re arbitrary with respect to the program itself, which is semantically empty. Being ‘all syntax and no semantics…” (Boden, 2018, p. 233).
Jaron Lanier basically agreed with Searle when he wrote, “Information is alienated experience…Information of the kind that purportedly wants to be free is nothing but a shadow of our own minds, and wants nothing on its own” (Lanier, 2011, p. 23). In short, information is not alive and does not have intentionality, therefore it can not have actual intelligence. I side firmly with Searle and Lanier that true intelligence presupposes understanding. I say, if it doesn’t understand, call it something else.
Section 2: What would Feminist AI look like?
A feminist AI would focus on aspects of AI that are currently lacking because of masculinist bias in AI research; Along with true emotional intelligence, there would be a focus on the ideas within the realm of ‘embodied cognition.’ Embodied cognition is a series of theories that claim that the body is an integral part of cognition, some theories also include the physical environment as part of the process of cognition.
Is affective computing a solution? Isn’t that emotional intelligence?
Emotional intelligence is defined as “the ability to understand the way people feel and react and to use this skill to make good judgments and to avoid or solve problems” (Cambridge, 2013). So-called AEI, artificial emotional intelligence or ‘affective computing’ just means the ability for a computer to predict a person’s emotion based on facial expression (Pantic, 2018). This still does not mean that it understands the emotion, as understanding in this case would imply empathizing. The problem again is the naming and overhyping of the abilities of the technology. These programs are not ‘emotionally intelligent,’ they’re simply recognizing patterns. Echoing back to Searle, they do not ‘understand’ emotions, therefore they have no ‘emotional intelligence.’
How is embodiment important to intelligence? Why?
The modern idea that cognition does not occur only in the mind, but is situated in a body and environment has become clear by combining evidence from many different fields, among them psychology, AI, cognitive science, neurobiology, and critical theory. There are currently a few different terms for these overlapping theories.
One of these theories is called ‘extended mind.’ Proponents of this theory have argued that AI is just part of an ‘extended mind’ in which “cognitive processes are not located exclusively inside the skin of cognizing organisms” (Rowlands, 1999, p. 22). So AI would be part of an ‘extended mind’ in which cognitive power is situated outside of the human mind and body. Similarly, ‘embodied cognition’ proponents like Margaret Wilson attest that “cognition is always situated” within an environment, therefore the environment becomes part of the cognition and can’t exist in an unsituated state (Wilson, 2002, p. 625).
Cybernetics based philosophers, like N. Katherine Hayles, argue that ‘embodiment’, being situated in a living body, within a natural and cultural, dynamically shifting environment is the only way to create genuine intelligence, as intelligence doesn’t situate only in the brain, but in the active dynamic between nature, culture, and body/brain. “And since, on this view, genuine intelligence is body-based, no on-screen AGI could really be intelligent” (Boden, 2018, p. 237). In summation: without embodiment there’s no intelligence, without intelligence there’s no understanding, without understanding there’s no empathy, and without empathy there’s no emotional intelligence. And these are all things that will be vital in order to approach anything like a General Artificial Intelligence.
It seems that the interdisciplinary nature of the human mind/body is beginning to be understood. All of these ‘extended’ and ‘embodied’ approaches represent a movement towards a more Deleuzian, postmodern understanding of the human as ‘assemblage.’ This acknowledgement of the complex interactions between systems both within and across mind, body and environement instantiate a more holisitic, feminist perspective on cognitive processes because it refocuses away from the narrow confines of reason and computation and towards a configuration that acknowledges the multiplicities of intelligence and perspectives.
CONCLUSION
‘Why Aren’t There More Women Futurists?’
Is the Singularity on it’s way? Are we hurtling towards a Mad Max dystopian nightmare? Or towards a virtual world where we’ve all uploaded our consciousness and are happily existing inside the Matrix? Or maybe just a world where all geeky men can fall in love and never have to feel alone or isolated anymore…until their AI girlfriend breaks up with them of course (you’ve seen HER I assume). You do have to pick one of these though. These are all of the options so far, at least according to pop culture and to most of those imagining our possible futures.
In an article for The Atlantic in 2015, Rose Eveleth asked the question, Why Aren’t There More Women Futurists?. She wondered, “Why can’t people imagine a future without falling into the sexist past? Why does the road ahead keep leading us back to a place that looks like the Tomorrowland of the 1950s?…The thing is: The futures that get imagined depend largely on the person or people doing the imagining” (Eveleth, 2015). But does it actually matter if most of the people imagining and building AI are men? It really does. Eveleth argues, and I agree, that “when only one type of person is engaged in asking key questions about a specialty — envisioning the future or otherwise — they miss entire frameworks for identifying and solving problems.” If AI continues to focus heavily on computational solutions it might stall, but worse than that would be it continuing down a gender unbalanced path that assumes one universal perspective is all that’s needed to build a balanced future for everyone.
I’ll admit, it is slightly disappointing to realize that you probably won’t be a part of the generation that figures it all out. We most likely won’t be uploading our consciousness and ‘living’ forever anytime soon. To remind ourselves that we’re still struggling with philosophical issues from the era of the Enlightenment (What is intelligence? What is consciousness? What is anything?) doesn’t make people feel like the revolutionary futurists they would like to think they are. But, honestly, after you get past the gloss and rhetoric, AI is, so far, just another tool for humans to use and mis-use. It’s a hammer. It can be used to build a home or to batter your neighbor. Evgeny Morozov echoes this sentiment, “…it’s much easier to proclaim yet another digital revolution — and to coin a requisite buzzword — than to wait and see if the observed change, instead of being a complete overthrow of established practices and principles, is just a shift in order and magnitude” (Morozov, 2014, p. 36). To admit that it’s still the same patriarchy and that technology didn’t throw off the shackles of gendered identity, does feel less exciting, but also comforting in a way. It’s just the same fight, different field.
Copyright 2020. Brighton Hudak-Kay