The Consensual Hallucination Is Here
Last year I was struggling to grasp what the coming onslaught of AI tools would mean for us. It was still early enough in their evolution their final form wasn’t clear. Actual use seemed limited and full of errors.
It's still early on in their evolution. And while the future is still relatively opaque, it is less so. We can start to see shapes and movement through gauzy curtains. The forms we see moving on the other side might be gods, they might be demons, or they might just be us.
The errors are still there, but they’ve been blended into our augmented reality so smoothly that you only notice them if they’re in your field. The R/Consulting subreddit is full of posts who say that AI will never be able to do their job in the same way they do it, and that it gives such basic stuff that it’s useless. The Marketing field is the same. And I’d imagine that if you ask almost anyone, they’ll admit that yeah, it’s good for general advice, but when it gets to the nitty-gritty specifics of THEIR field it’s confidently wrong. It can help you get the basics, but you need an extra to make it work.
In Neuromancer, the godfather of Cyberpunk and unwilling visionary of our current era, referred to the internet as a consensual hallucination:
“Cyberspace. A consensual hallucination experienced daily by billions of legitimate operators, in every nation, by children being taught mathematical concepts... A graphic representation of data abstracted from the banks of every computer in the human system. Unthinkable complexity. Lines of light ranged in the nonspace of the mind, clusters and constellations of data. Like city lights, receding...”
Neuromancer is a good book. The writing and technology in it generally hold up well. And it asks the questions we should be asking about General Artificial Intelligence, that spectre haunting the singularity horizon. It was a few years away when he wrote Neuromancer, was a few years away ten years ago, and will likely be a few years away next year and this year as well..
The cyberspace he wrote about and described, and that of the early internet, wasn’t a hallucination at all: the data and complexity was at least grounded in reality and fact. It was the opposite.
What we’re seeing with generative AI is a better example of Consensual Hallucination. A collective agreement that what these tools make and output and tell us is true and real, even when it isn’t. AI influencers, AI porn stars, AI relationships, AI friends, AI social media feeds, AI journalists, AI movies, and AI musicians.. We’re agreeing that this is normal, expected, and inevitable even as we acknowledge how fucked it is.
All of my earlier thinking about AI was firmly about how strong and capable the ‘I’ aspect of it would be, and how blurred the line between the A-for-Artificial and the human “us” would be.
I ignored the generative aspect, but now it’s gotten me thinking about a few other Sci-Fi books that will help us understand the present day. Specifically, House of Suns and Rainbows End.
House Of Suns by Alastair Reynolds might be the best big-concept sci-fi book ever written. It has near-immortal clones that share memories by dreaming. It has enigmatic and handsome robots. It has a galaxy-wide mystery. It has chase scenes that last hundreds of thousands of years across space. It has the best kind of ‘Distant Mountain’ world-building that you can imagine, even just in the form of the
Shuffled into those operatic pages is another story entirely, told in parallel with the main one. It’s the story of a young girl named Abigail. She lives in a sprawling, labyrinthine structure on an airless moon, the daughter of an exceedingly wealthy family. Her mother lies in some sort of life-support stasis and she is cared for mainly by robots.
One day a group of specialists arrive and install a new device: Palatial. I don’t think it’s ever referred to directly as a game, but it’s essentially a holodeck-type of immersive-VR device that takes you to a fantasy world of knights and kings and damsels and maids-in-waiting.
Abigail and a young boy who visits her become immersed in the world of Palatial, to the point of obsession. It feeds on their subconscious and memories, and continues to build out a narrative for them.
It’s as real to them as real life, and it shapes how they think about each other and their world. I don’t want to spoil much more than that…except to say how truly in-depth the Palatial world is.
I believe that one future for us is this: individually immersive worlds that cater to our whims and wishes, and don’t require us to interact with a world or people outside of those virtual worlds. The hallucination becomes everything.
*I predict that we will see at least one AI-company called Palatial within the next few years, based solely on this and non-ironically referencing the evergreen “Torment Nexus” tweet.
Rainbows End by Vernor Vinge presents a different view of generative AI, though also likely influenced by video games. Via in-eye augmented reality, people are able to give the world they inhabit overlays to depict their own interests. For example, my interest in the middle ages might result in me viewing our world as a medieval fantasy: the walls of apartment brick and stone, the fireplace real, the posters on the wall woven tapestries instead of family photos. You could co-inhabit this same space, with a remarkably different overlay, say one of a space station. We’d be able to interact with each other, and with the physical objects in the world. Our reality would be different.
The result is in some ways what we’re already seeing with our different views of reality based on our algorithm or the social networks we choose to inhabit: a fracturing of our world view. A lack of central framework for even the most basic things.
One of the things I’ve consistently said about why I like science-fiction is because it helps me understand the way new technology or inventions will influence us. Books like this give us a chance to ask “what if…?” and explore it not just from a product roadmap and features standpoint but from the way they will shape and influence society. These books will often get a lot of things wrong, and that’s okay. The point of reading is to broaden your thinking. For it to be a branching pathway to different avenues of thought, and ideas. I feel like I’m better prepared for the future because of what I’ve read. And as I’ve said: I’m optimistically terrified about what’s next.
I watched the movie Casino a few weeks ago, and couldn’t help but notice how nearly the entire cast was smoking cigarettes throughout the movie. It was a product of the times, I suppose: A movie shot in 1994, about the late 1970s and 1980s. I’m sure at the time, you could buy cigarettes on the casino floor, from a desk nearby, or from actual vending machines.
As much as I noticed the cigarettes, I also noticed that Robert Deniro, as casino manager Ace Rothstein, used a cigarette filter. He might have been aware of the dangers of smoking, but was unable to stop for either physical or social reasons. Or he might have just liked the affectation of it.
I feel like we’re in the “cigarette”-era of generative AI. It’s socially acceptable, and even encouraged. But we’re also starting to be aware of the dangers. Some of us are leaning in and smoking more. Others are leaning out. I’ve got my filter on. I’m not sure it will do much.
I spend a lot of time thinking about Science-Fiction. For me, it’s both an escape and an intellectual exercise in trying to make sense of the future and technology.
If you liked this article you might like two others:
Optimistic Terror: Sci-Fi To Read To Understand Artificial Intelligence