Last weekend, I was able to participate in a great event. It was put together by Storyline Partners and Luminate at the very cool and welcoming Blackbird House in Culver City. It was called Writing in the Digital Now and it brought together a group of writers to hear from two academics in emerging technology and society, Dr. Nishant Shah and Dr. Maya Indira Ganesh. It was a particularly exciting and informative workshop that really blew my mind.
Now, when you hear “a conference about new tech,” you’re probably thinking about AI/LLM/machine learning and probably concerned that the subtitle of this event is “How to stop worrying and learn to love AI.” You wouldn’t be alone. Even some of the writers in the room had those concerns. But that really couldn’t have been further from the truth (at least from where I was sitting).
The conference was really about narratives. About the ways we talk about tech. About how tech of all kinds and stripes is represented in the media. And how writers can change those narratives, many of which we helped to create. As a genre writer with a strong interest in science fiction and speculative fiction, the whole thing rustled my jimmies (as a friend of mine is wont to say). Because thinking about how we think about tech is a big part of my job.
One of the big things I took away from it is how much, when we’re talking about tech (and science, in general), we use metaphors. “The computer is a brain” is a metaphor. “Schrodinger’s Cat” is a metaphor. “The Amazon rainforest is the Earth’s respiratory system” is a metaphor. And the thing about metaphors is that they’re not precise or even close representations of reality. They’re a way to process and understand something that is, in reality, much more complex, much more interesting, much more difficult to process.
But they also get locked into our language. We use terms for the brain, for biology to describe computers and vice versa. We reinforce those connections constantly. So much so that we stop thinking of them as metaphors. We think they accurately represent reality. But they don’t.
Obviously, one of the clearest ways we see that now is in the dialogue around Large Language Models and machine learning. We call it Artificial Intelligence. We talk about hallucinations. We talk about the various chatbots having personalities. We say that they’re thinking. We worry about them becoming sentient. We infuse them with agency and opinions and potency.
But that’s all metaphorical. The chatbots aren’t doing any of those things. They’re not thinking machines. They’re algorithmic probability code. They’re not even really “machines’ in the way we think of machines. They’re code that scans text or images, give the individual words, characters and images a numerical value, a weight and then re-assembles them in the most probable order. And even that is probably not an accurate metaphor. Because the devil is in the code, the underlying maths that I assure you I can not comprehend.
Metaphors are, of course, useful for communication. We’re not all computer scientists. But metaphors can also become a trap. Because they’re drawn from our own understandings, they’re drawing on our ways of seeing the world. With that comes a lot of deeply encoded ideas about how the world works. Ideas that are drawn from our society, our culture. Why do you think we talk about masters and slaves when we’re talking about computers being paired together?
And what does that do to our understanding when we think about them that way? That was a big part of the conference. Getting us to understand that how we think about the world bleeds into our metaphors and then that shapes the narrative about what a tech is.
There’s an idea that I’ve come across a few times that’s usually phrased as The Terminator Problem. It’s the idea that our ideas about tech are shaped by the media we’ve seen, even when that media is not presenting an accurate picture of the tech or even where we are, technologically. It’s called The Terminator Problem because, whenever we talk about AI, invariably Skynet, the all-powerful, human-hating, fully sentient artificial intelligence from that franchise enters the chat. But what we call "AI” has almost no actual relationship to what Skynet is. Not just in terms of orders of magnitude, but in terms of actual infrastructure, design and use. A lot of our fears about the possibilities of Large Language Model technology are really fears of a wholly fictional boogeyman that doesn’t actually exist.
Again: this is not “how I stopped worrying and learned to love AI.” I’ll get into what I find so pernicious about the narratives around LLMs, chatbots and that entire ecosystem in a second. What I’m saying in this part is that our inability to separate the narrative from the reality makes it harder to address the actual reality. It makes it harder to legislate, to regulate, to understand. We’re responding to a metaphor instead to an actual thing.
For a long time, I worked on The EST Sloan Project, a program to expand the public’s knowledge and understanding of science and technology. That mandate meant avoiding science fiction and focusing on the most current scientific research, as well as the history of science and scientists. For me, as program staff, it meant talking to a lot of working scientists and science writers and educators, learning the limits of what could actually happen in science and the best ways to depict that clearly. I learned a lot about the history of science and the lives of various scientists and also the latest advances in a lot of different fields. Let me tell you, once you start learning about what’s actually possible, what’s actually happening, it kind of ruins a lot of science fiction. Science fiction, for the most part, runs on metaphor and what if and depicts things that are not only far beyond our current capacities, but also, frankly, utterly impossible.
A recent example of that is the very excellent film Project Hail Mary. Great, thrilling, emotional movie. But, boy howdy, do they hand-wave past a lot of very tricky science. Like how a faster-than-light ship was developed and built. And how it was so easy for Grace and Rocky to communicate. None of these are fatal errors; the movie isn’t about the tech or those specific issues (unlike a movie like Arrival). But it leaves us thinking that those things are easy and possible. When they’re not.
In a way, that kind of thinking also undercuts our understanding of how complex and amazing our own biology is. For me, the “my brain is a computer” analogy breaks down when I think about one of the simplest things I do every day: crossing the street. I live near a wide, four-lane street with a median in a city where people drive as fast as humanly possible at every chance. I often jaywalk across that road, to go to stores, cafes, whatever. And when I break down what that requires into its component parts, it’s actually amazing. With my human eyes and my human brain, I can calculate how far away a car is, how fast it’s going, how far the other side of the road is, how quickly the car will reach the spot where I am, how fast I need to move to get to where I want to go, all in micro-seconds. And, at the same time, I’m processing the air temperature on my skin, the feeling of my shoes on my feet, the angle and intensity of the sunlight, while adjusting my walking speed to account for all of those factors. While also thinking about my shopping list or my coffee order or my to-do list. While also thinking about that song I like, that embarrassing thing I said ten years, the last thirty times I made this crossing and why the hell is that car driving so fast. Not to mention the parts of my brain that are functioning automatically to keep my heart beating, my lungs inhaling and exhaling, my muscles contracting and releasing. That’s so many layers of processing, all happening at once and most of them without any conscious thought. When I think about it that way, it’s a magical thing that we do. And we just…do it. Several times a day.
Of course computers can do many of those calculations and a lot of them do. Another scientist I’m a big fan of, Dr. Rachel Barr, recently posted about the ways in which our brains and LLMs do similar processing. We’re both scanning the environment, taking in information, assigning value and weighing probabilities. But that still doesn’t mean the two things are like each other. Planes and cars do many of the same things, but they are not the same. And that’s a good thing.
There’s another thing about these narratives. They don’t come from nowhere. Now we get into the meat of the thing. I have no actual feelings about Large Language Models, about machine learning, about probabilistic programming. I do have very strong feelings about AI and about the narratives that are being increasingly embedded in our society, driving the conversation around it. Because those narratives have been created, not by technicians or scientists, but by investors, by salesmen. It’s in Sam Altman’s best interest to continue to feed the narrative that ChatGPT can do our thinking for us, that it’s the most powerful, life-changing tool in the history of humanity. It’s even in his best interest to keep us scared about AI taking over the world. Because that means we give him money to protect us from it.
These narratives didn’t just well up and don’t propagate themselves. They’re pushed, shaped and controlled by the people who directly benefit from them. So, yes, once again, the villain is actually capitalism. That’s the part that I don’t trust. It’s become clearer and clearer that these chatbots can’t actually do any of the things that their sellers claim they can do. It’s a grift. But once a narrative gets set, it’s hard to unwrite it.
That, to me, is ultimately what Writing in the Digital Now was all about: finding ways to overwrite this programming (to borrow a metaphor) and finding ways to remind ourselves and everyone else that these narratives aren’t immutable facts. They’re stories we tell ourselves. and we can find other stories to tell, other ways to shape the story, other ways to communicate. Imagine if we stopped saying that ChatGPT hallucinates and instead said it throws errors. Reminding ourselves that there’s no brain there, just code. And when code makes a mistake, we call it an error. We call it a bug. If we remind ourselves that these are programs that have been written, we can remember that someone is writing them. Someone is making them. A lot of our issues are not with the chatbots but with the very human people writing them with all of their human self-interest, intentions, assumptions and limitations.
Those were my biggest takeaways from the weekend: we are in control of the narratives and we can rewrite them. We can re-think them. We can craft new narratives, new metaphors that center different ideas. As a writer, I take it as part of my mission in this world to imagine other worlds, other ways of existing, other ways of solving problems. It was good to be reminded of that. To be reminded that, as a metaphor, the world is a story. And stories can be revised.