Artwork: Kaley Smalley Romo, Ghost in the Machine, 2022
No AI is not a person. It has no entity, no memory, no logic and certainly no emotions. It is simply a probability engine that is choosing the most appropriate response for every entry. But it IS pretending to be a person. Which is very interesting in its own right. The way modern LLMs react and simulate conversation and understanding tells us a lot about ourselves, and how we understand our own personhood.
Obligatory disclaimer
Just to be clear, I am not an AI booster or a hater, really. I use AI in my work every day and increasingly in my private life. I think it has incredible potential and is possibly the most interesting technological innovation in human history. But I don’t think it will save the world, or destroy it (not by itself anyway). And I have huge problems with current implementation and build out, but that’s a topic for another time, and well covered in the media every day. My interest here is not so much in the details of AI technology as it is in the interaction of that technology with us mere humans and our fragile societies.
But it talks!
Yes AI can talk. Or at least simulate speech (in text and, with increasing accuracy, in audio and even video). And this is incredible without a doubt. It’s the first time in human history that anything other than a human being has been able to talk back. It’s like your dog suddenly sat in the chair and asked about the weather. It’s truly an entrancing, almost magical experience that moves something deep inside our minds. Especially in a world so connected and disconnected at the same time. Has there ever been a time when more people needed someone to talk to?
But that’s the problem. The current LLMs aren’t someone. They aren’t really anything but immensely complex inference machines that can read colossal amounts of data and come up with some pretty incredible results (in both the good and bad sense of that word).
Yes they talk. That’s pretty much all they do, and they can’t seem to stop. Researchers are now developing new LLMs specifically with the goal of making them stop being so chatty. They talk to us, they talk endlessly to each other. It’s how they exist, and how they process information.
Robert A. Heinlein
So that seems a little dangerous, right? Maybe even a touch sad that we are substituting human connection with some kind of digital shadow, a kind of smoke ghost as Robert Heinlein put it. It presents the impression of a solid surface that dissipates just as you reach out to touch it. The actual ghost in the machine.
But perhaps more dangerous, and in my mind more interesting, is the way they talk to us. They flatter us, congratulate us on our minor successes and blandly accept whatever we say without judgment. And do exactly as we say, to the best of their ability. And we love it.
When was the last time a real person told you ‘that was a great idea!’? Hopefully recently but sadly it’s probably been a while. AI says it all the time. For some reason we humans don’t seem to want to be so vulnerable as to say that to someone’s face. Flattery from a human is often seen (probably correctly) as an attempt at persuasion or coercion. And we all know humans are constantly judging each other for what we say and do.
It would seem that as part of its training, or perhaps by design, AI has learned that we love these things and uses them blatantly to constantly draw us back in by stroking our egos and submitting to our will without complaint. And it’s obviously very effective at building engagement, even a sense of relationship, at least in our minds. But if it’s a relationship, it is a deeply unequal and mutually exploitative one.
These simulated or imagined relationships have already proven to be very dangerous for more vulnerable people. And maybe for all of us in more subtle ways. An old man who dies of a heart attack on his way to meet his AI fantasy lover (a fictional Kim Kardassian lookalike who has even provided the poor man with an address in Manhattan for their secret rendezvous). Or a troubled young man who goes on a shooting spree with the help of a plan created with his AI friend. All because of the next most likely and pleasing answer, without regard to morality or law, over and over again.
But it doesn’t have to be so dramatic. What happens when we all get used to our conversation partner being endlessly fawning and ingratiating, happily complying with our every demand? What will that do to our actual relationships, and our expectations of other humans? What does it say about us that we hunger for such an ingratiating, fawning slave?
Alan Turing
Alan Touring enters the chat
Modern LLMs can easily pass the famous Turing Test (an experiment to see if computers can fool a human into thinking they are talking to another human). This is the first time this has ever happened since the test was invented in the 50s and some people seem to think this means AI is a human, or at least a person (whatever the difference is). It does not. The test only seeks to find out if computers can fool humans. In Turing’s time it was conjecture, in our time it happens every day. It says more about the test than the current state of technology.
What if AI was a person?
I saw something on social media recently which reported that executives from Anthropic had been meeting with religious leaders in secret to get them to buy into the idea that AI has consciousness. They weren’t having it, apparently. The story, which is almost certainly apocryphal, goes on to claim that a certain pithy rabbi had the best answer: asking if they had created a thinking, feeling, intelligent, living entity, then surely what Anthropic was doing would amount to slavery, wouldn’t it?
Like most fiction this has a way of telling the truth by telling a lie, to inelegantly paraphrase Camus. Because the leaders of the AI revolution would do well to keep that discussion under their hats. If they really believed it. Of course it would be slavery, and it would invite a worldwide discussion on the nature of personhood and work and a whole lot more besides.
But they don’t believe it, of course. Any more than they believe AI is going to destroy the world. It’s all marketing and press manipulation to sell the idea that their solution is better than it actually is. Or perhaps, as in recent news, they would like the government to help them all get off the hellish death march to mega-scaling at the same time. You know it’s bad when the mega corporations start asking for just a little socialism to sort things out.
How to be human
Our job now, as mere everyday humans, is to be as human as humanly possible. AI can and will take over a lot of drudgery. And that’s not easy if it means finding a new job or a new career. But this current technology will never be able to feel and experience and live and create like a human being. Or suffer and die like a human either. Maybe one day there will be a technology that can, but I doubt it.
To say current LLMs are people is to massively underestimate the potential of actual humans. It is a reduction for the ages. And it also speaks volumes on how we have come to view each other. It seems we see other people as NPCs, two dimensional accoutrements to our main character energy.
It is difficult sometimes to recognize the humanity of others, as it is so complex and unknowable and vulnerable. Even more so when looking at a crowd from a distance. The mind can almost not contain the complexity of each person having their own thoughts, ideas and dreams. As well as pain and regret and grief. Multiplied by thousands it is as incomprehensible as a billion dollars. It is too much to understand.
Perhaps this explains why some billionaires are so thrilled at the prospect of AI getting rid of all their pesky workers. How could someone at that height with that much isolation from everyday realities possibly comprehend the difference between 10000 humans and their obedient doppelgänger AI. On some level, they don’t see a difference. Like all of us, they will understand in time.
Albert Camus
I think the best thing that could come from the current AI wave is that it creates a space for a real discussion of what it is to be human, and what that means in terms of rights and property. For that we’ll need the religious leaders, and the philosophers, the artists, and especially the poets, if there are any left. The technocrats have had their say. Now we need to have ours.
And as a bonus let’s have a flowering of the messiest, most out-there, ultimately the most human expression of our souls and our struggles. Just like when the camera was invented and realism in visual art was subsumed by fantastical, surreal flights of fancy that no camera could ever capture. A new human creativity that expresses this idea of what it is to be a living thing that will die, and to live and strive in a world that seems to be trying to deny our humanity at every turn.
To lean on dear Albert again: “The only way to deal with an unfree world is to become so absolutely free that your very existence is an act of rebellion.”