AI has being but it does not exist – an existential approach to the difference between humans and AI
In the last few weeks Warner Music Group and Disney have partnered with AI firms, and this has left many to fret about “losing human creativity” to AI.
We need to be very careful how we frame “human creativity”, because as scholarship in a range of fields has shown, any time you try to define “the human” in terms of essential qualities or capacities, it’s impossible to fully capture all of the actually living persons that we would commonsensically call human, such as disabled and impaired people, intersex people, people practicing non-Western cultures and their values, just to name a few of the most common exclusions. If you say the thing that defines humans is “language use,” then human babies and some disabled people don’t count but non-human primates who can use sign language do. For the same reason, it won’t work to try to find some objective property of human creativity that distinguishes it from AI. Human creativity is so diverse and varied that it even, in for example the case of Thomas Kinkade, basically mirrors AI slop. Compare Kinkade’s “Rosebud Cottage”
To this AI-generated image on Freepik
We can assume that Kinkade’s body of work has been used to train these image generators, but the point is Kinkade came up with this cheesy stuff all on his own! With his human brain and creativity! There is a lot of human-made slop out there, some of it, like Kinkade’s, is even extremely practiced and refined.
When it comes to distinguishing between “humans” and AI, existential phenomenology offers what I think is the most helpful and accurate toolkit. Specifically, I think Beauvoir’s distinction between “being” and “existence” can distinguish between people and artificially intelligent machines: AI has being, but it does not exist because it’s not a mutual participant in the co-creation of existence.
As an existentialist, Beauvoir thinks that humans have no essential being, i.e., that human “being is lack of being” (13); that’s her way of saying something like “existence before essence (aka being).” Because I have no essential being, I can shape my existence in any way my situation facilitates. For example, human bodies can’t fly like birds or swim like fish, but by working together we figured out how to build airplanes and SCUBA diving equipment; we considered the material affordances and limitations of typical human bodies and figured out how to make them fly and dive. Rather than treating the facts of typical human embodiment as an essential, fixed form of being, this perspective understands these facts as contingent situations that can be worked with and through. This is what Beauvoir means by the idea that humans “lack being”: there is no singular, universal, fixed, eternal, unchangeable human reality. However, humans are real: we exist. And our existence’s contingency and malleability is possible because we lack being. As Beauvoir puts it,
this lack has a way of being which is precisely existence. In Hegelian terms it might be said that we have here a negation of the negation by which the positive is re-established. Man makes himself a lack, but he can deny the lack as lack and affirm himself as a positive existence. He then assumes the failure [of being]…However, rather than being a Hegelian act of surpassing it is a matter of a conversion. For in Hegel the surpassed terms are preserved only as abstract moments, whereas we consider that existence still remains a negativity in the positive affirmation of itself. And it does not appear, in its turn, as the term of a further synthesis. The failure is not surpassed, but assumed” (13)
Note the distinction Beauvoir makes between “surpassing” and “assuming” a lack of being. To surpass or sublate (Aufheben) the lack of being is to cancel out the tension between being and nothing by positing another form or level of being. On the other hand, to assume the lack of being is to acknowledge it and not cancel it out; to exist is to have a reality that is not being. Existing (an active process) is not being (a static state). In Beauvoir’s system, existing consists in doing things in the world, changing and re-orienting ourselves and our situations. In so doing, i.e., in existing, we negate a present state of being and create another one. An affirmation of a negative, assuming one’s lack of being consists in acting upon the fact that one has no inherent, immutable reality by changing one’s present situation and creating a (slightly) new reality. Whereas Hegel and most of the Western philosophical tradition take “lack of being” as a bug in need of fixing or resolution in some higher order of reality, Beauvoir considers it to be the defining feature of human existence: existing reorients a present state of being into something else.
In claiming that humans “lack being,” Beauvoir’s model cannot be made to endorse or frame any particular mode of being as more legitimate than another. Being is there to be negated in acts of existing, which reorient the world in new ways. What matters is existing in ways that are oriented toward…existence itself, or the act of negating a present state of being and creating a newly-oriented one. Because individual existences are enmeshed with one another, orienting my actions to the diminishment of others’ existence also diminishes my own, and orienting my actions toward my personal existence also orients them to the existences of others. In the Beauvoir scholarship, this orientation is called “moral freedom.” The language of freedom points to the non-essentialist character of existence: lacking being, existence has no necessary features or determinations limiting its future directions. This freedom is moral, however, because human existence is, in Beauvoir’s ontology, inescapably relational: my own freedom depends on the capacities of others to exercise theirs. As Arp notes, “If one acts to curtail others’s freedom then one cannot achieve moral freedom” (28); a freedom that fails to orient itself to the freedom of others fails to be moral not because it fails to meet some external moral standard, but because it fails to fully orient itself toward freedom/existence. Beauvoirian existence cannot itself be taken as a legitimating ground because it has no being to legitimate. Framing existence as ambiguously both individual and common, Beauvoir’s ontology creates a system of accountability that can operate in the absence of legitimating values such as “the greatest good” or “virtue” or “life.”
Our actions orient the world that we share with other people. Because each action is oriented more positively toward some and more negatively toward others, we are accountable to all in our individual choices. For Beauvoir, oppression prevents people from genuinely assuming their existence, i.e., from being able to negate a present state of being and create a new one. For example, recent attempts by U.S. higher education institutions to criminalize student protest impede these students’ ability to have a say in what sorts of things their institutions invest their money in, such as weapons manufacturers. These protests are an attempted exercise in moral freedom: they are actions in service of negating a present reality (that their school sends money to make weapons of war) and create a new one oriented less towards the military-industrial complex and more towards decoloniality. By criminalizing this protest, U.S. colleges and universities prevent their students from contributing the exercise of their moral freedom to the mutual orientation of humanity’s existence. Oppression diminishes human existence at both the individual and the general scale. Thus, because Beauvoir’s existentialist ethics holds that there are no absolute, universal values beyond the fact of human existence itself, oppression is uniquely condemnable because it is oriented directly against existence.
Human existence is collaborative and mutual: my existence is oriented to and oriented by others’, and we are all responsible to one another for our contributions to our shared reality. AI cannot take responsibility for its outputs. People respond to those outputs, and in this respect AI can impact our shared existence, but only insofar as it inspires someone to do something. AI doesn’t have the moral freedom that humans have. And this moral freedom is rooted in our social and political orientation to one another; the thing that humans have and AI does not is society. In other words, the difference between humans and AI is not biological or cognitive or related to any objective feature or characteristics of either us or them; it’s a political difference rooted in AI’s inability to be responsible and thus be in community. To the extent that people start to rely on things like chatbots rather than on one another, AI is a form of bad faith that actively erodes human existence, i.e., society.
I don’t think AI meets Beauvoir’s other criterion for existence: it is not capable of negating a present reality and creating a new one. First, it can’t do anything actually new or original. It generates probabilities based on patterns in past data. It can recombine its inputs, but it cannot negate those inputs and put out something different from them. For example, I’m sitting here at my laptop writing, but when I get up to let the dogs out I negate my present reality and push my reality along another vector. AI can’t do that: if all it had in its training set was sitting at the laptop writing, it could just come up with iterative variations on that. Unless it had some data about standing up to let the dog out in its set, it’s not going to suddenly generate that as an output.
Because AI can’t convert being into existence by negating its present reality and creating a new one, it lacks the capacity to participate in the mutual creation of existence. Like Beauvoir’s figure of the child, AI-resurrected grandma “escapes the anguish of freedom…her whims and her faults concern only her. They do not weigh upon the earth. They cannot make a dent in the serene order of a world which existed before her, without her, and where she is in a state of security by virtue of her very insignificance. She can do with impunity whatever she likes. She knows that nothing can ever happen through her; everything is already given; her acts engage nothing, not even herself.” Just as children playing house aren’t making decisions or actions that impact anything outside that play scenario, AI-resurrected grandma lives in a playspace separate from the messy real world of human existence. If living people act on her advice, they face the responsibility and consequences for acting on information generated in a play-world.
AI may function in a play world, but the machines that run it have a very real material impact on the world. But that impact is not the result of the AI’s choices – it’s the result of all the companies deciding to build huge server farms that gulp up all our water, much in the same way that cows who fart methane into the atmosphere at quantities to impact greenhouse gasses aren’t responsible for their existence at that scale, agribusiness is.
We are bags of meat stuck here on this rock together – and it’s that together that defines us. That’s a political and social definition, one grounded in moral freedom. AI does not share our reality because it does not exist together with us – it just has being.