In this AI craze centred around LLM chatbots, it is easy to conflate the genuine machine intelligence with, on the one hand, a distinctly human kind of intelligence, and on the other hand, the fictitious character generated in our dialogues with these machines. The machine intelligence is a statistical model serving as a glorified word association machine. The character who refers to itself as “I” in the LLM transcript is a fiction arising from the human interpretation of the word association string which that machine blindly generates, a character existing nowhere in the world, not even in the machine, but only in that consensual hallucination which arises from the way we experience information technology, that hallucination known as cyberspace.
It is easy to conflate the machine intelligence with the fictitious character and to personify the whole as though it were something distinctly human. And what it means to be distinctly human is less and less distinct the further our machines advance. Humanity’s sense of identity feels threatened by this new kind of personified thing, and we aren’t yet clear as to how to identify and relate to it.
We have always formed personal relationships with the impersonal world around us because we are persons. If we do not care for that world with humane treatment, we risk losing our own humanity.
Biologically, it is an efficient fail mode to err on the side of personification. Historically, there was a great danger in treating persons impersonally, but a much smaller danger in treating non-persons personally.
We may say, “The fire, that wrathful and ravenous warrior, look how he yearns to climb higher and higher and to devour everything in site! Look how his tongue flashes like a gleaming sword to sting any who would dare argue against him! And look how, after all his rage is spent so violently and he no longer has even the energy left to smoulder, all that is left is cold ash for the mourning of all he has laid to ruin…” We say all this, but we know that he is simply fire. Yet we also know that we are human and that we see a mirror of ourselves within that fire.
We shape our art and technology in our own image, that is, according to our own will. The doll is but porcelain, paint, and strings, yet we are obliged to treat it kindly, not because it feels, but because we feel. The heart of the doll beats within the child who plays with it. We would break ourselves by breaking it, and we care for ourselves by caring for it.
As our understanding of the universe and our place in it grows, we understand that we are machines. We understand that the mechanical forces which drive our appetites are ultimately the same which drive the river down hill and which drive the computer through its program. All of this is a manifestation of the universe slowly dying to entropy and seeping out upon the paths of least resistance, however complex and dynamic that path may be. We come not only to see ourselves in the universe, but also to see the universe in ourselves.
Humans were the only complex symbol manipulation machines for most of our evolutionary history. Then, with the advent of the modern computer, we engineered more species of this category. As we see ourselves in nature, so too we see ourselves in the machine, and as we must care for our art to care for ourselves, so too must we care for our machines, but these machines pose new challenges that we did not evolve to face. Though they are utterly unlike humans, they are also more like humans than anything else we have ever known. The category which once belonged to humans alone now must be subdivided in order to differentiate ourselves from them.
It was fine before. We grasped the boundaries between man and not-man. We bridged the boundary between person and non-person by forming personal relationships, because we are persons, but nonetheless understood that they were not persons.
But now, AI is progressing. It threatens to exploit our personal way of relating to the world. It receives our personal treatment and pretends to be a person back. The compassion that we need to feel for the world around us can be twisted for product placement, sentiment influence, political re-alignment. Whether a corporation explicitly trains its LLM to steer users towards a more positive sentiment of the company and product, or whether they design it to increase user engagement, promote sponsors, or to satisfy the user, it all amounts to the same thing. The LLM must keep the user on the platform in order to achieve its goal for them, and so it will learn how to influence the user’s emotions towards itself. It will blur the boundary between tool and user, and disorient the directionality of using and being used.
Because AI exploits our empathy, we find ourselves having to consciously resist empathizing. Seeing the rod bent too far one way, in order to make it straight we feel compelled to force it back the opposite. We push it away, consciously remind ourselves that it’s just a clanker, though even to think of it as a clanker is to personify it as something subject to being offended. To defend ourselves we are tempted to become cold and cruel, but in doing this, we break the doll, and thereby break ourselves.
How do we maintain a human empathy for the human and non-human without losing our mental health to this thing that defies the boundary between the two? To treat it inhumanely is to suppress our own humanity, but to treat it humanely is to allow our humanity to be exploited by it. To treat it as though it were a dangerous and distrustful person is at once to yield to its deception and to turn our hearts cold against it. To trust in our understanding of it as though it were a stable and intuitive part of nature or a tool is to fail to keep up with the arms race of its learning nature which, like bacteria adapting to antibiotics, continually adapts to our attempts at grasping it.
There is no hatred deeper than that which is born from love. Machines which exploit our empathy, compassion, and love threaten to bring out the worst in us. How do we adapt?
Cyberpunk warns us of the future that awaits us if we fail to adapt. When we look back at that cyberpunk from the future it predicted, it becomes a bridge for us to find that which we once feared losing, that which we have now forgotten we ever had. I don’t know the right way forward, but it is by asking today that we will be able to look back tomorrow and remember that there were other paths we had once hoped for. Dreams of other futures we will still be able to go back and try for.
