One of the masters of so-called, “hard” science fiction is Arthur C. Clarke. With degrees in mathematics and physics, Clarke knew science well. Several of his classics, such as Rendezvous with Rama and 2001: a Space Odyssey, contain lots of genuine science. When he was writing 2001: a Space Odyssey, he was very aware that real science, in the form of an upcoming moon landing, was snapping at his heels, in the sense that he had to make sure that the science he used in his stories was compatible with what was expected to be found on the moon. I’ve felt the same pressure.
In my novels, Ezekiel’s Brain and its sequel, Prime Directive, which will come out soon, there are at least three areas where scientific progress is hot on the heels of the concepts I use in my books. The first of these is in artificial intelligence, particularly conscious artificial intelligence. The second is in space flight, and the third is in knowledge of the stars and planets.
Ezekiel’s Brain, came out at the end of 2020, and ChatGPT was released to the world in November of 2022. All of a sudden, everyone was talking about AI, wondering if it were truly intelligent, and if it could become conscious. In the last few months, led by such such computer science luminaries as Geoffrey Hinton and Sam Altman, as well as philosophers such as David Chalmers and Daniel Dennett, the subject has been raised of whether AIs, such as ChatGPT, could become artificial general intelligence—AGI—and threaten humanity. That’s exactly what happens in Ezekiel’s Brain.
So, how is real science used in Ezekiel’s Brain? In the book, Ezekiel Job is a neuroscientist who scans his own brain, then reassembles the images into a 3-D map and uses a 3-D printer to duplicate it in silicon. He’s created what Oxford philosopher and AI scientist, Nick Bostrom calls an “emulation.” There are some profound difficulties with the actually doing such a thing, but Bostrom predicts it will happen someday, and, by definition, the result will be an AGI, an AI with the capabilities of a human brain. The problems are mainly two-fold: first, we aren’t even near the point of being able to scan and assemble such a thing, and second, unless it’s possible to capture the dynamic nature of synaptic weighting, the thing that determines which neural connections fire together as networks, and under what circumstances, all we would have would be an empty architecture, like the frame of a building but with no one using it.
Ezekiel’s Brain includes another AGI, one more likely to emerge sooner than an emulation. In the novel this AI is called Wanderer. Its creators assemble Wanderer by combining three special purpose AIs, all of which would have been developed using artificial neural networks and machine learning. One of these is expert at acquiring resources, and is modeled upon IBM’s Watson. Another is a Google-type search engine, and the third is a social chat engine. In order to coordinate these three special purpose AIs, Wanderer’s creators need to install a consciousness module, which has three functions: it can send and receive information from any or all of the three special processors; it organizes those processors to work together to carry out a task, and finally, it describes its goals and strategies in natural language so what it’s doing can be understood by humans and commands can be given to it in natural language.
I didn’t make up the model for Wanderer; a neuroscientist named Bernard Baars did, and he called the conscious part of his model, the Global Workspace. Baars recognized that the vast bulk of what our brains do is done unconsciously and automatically. When I speak to you, I am not aware of holding my head upright, of selecting words in a particular order to express my ideas, of organizing my mouth movements to say those words, or a host of other automatic functions that my brain orchestrates. What I am aware of is a general goal of conveying an idea to you through my speech. Even that goal is not prominent in my mind most of the time that I am talking. I’m only conscious of a fraction of what is going on in my brain, but that fraction of consciousness allows me to martial the help of all of those unconscious automatic processes to carry out my plan to reach my goal. This is what the creators of Wanderer had to add to their specialized processors to make their AI into a conscious, functioning artificial brain.
In Baars model, a necessary component is something he calls a “frame,” which embodies the perspective of the self as the actor who receives sensory input as perceptions and carries out actions. Consciousness usually operates within this frame and not doing so only occurs in abnormal conditions such as depersonalization or dissociation. In Ezekiel’s Brain, I made the assumption that the self-frame is genetically based, not learned, and therefore it had to be programmed into Wanderer’s conscious mind.
New findings with AIs based on artificial neural networks and machine learning algorithms have raised some questions about how consciousness and self-consciousness develop. A recent series of studies found that an AI that learned to carry out simple movements by imitating human movements could solve problems better it if was exposed to humans voicing the thought processes that guided their behavior. The AI then used verbalized thinking to decide how to solve a problem by combining different movements to achieve the solution. It outperformed AIs that only imitated the physical behavior and was better at solving novel problems to which it had not been previously exposed. This may suggest that the AI utilized conscious reasoning much as Baars’ Global Workspace theory predicts humans do. But did it have a sense of itself doing these things?
“Theory of Mind,” or as it is now called, “Mindreading,” is the ability of humans, above a certain age, to predict how people will behave based on assumptions about what they are thinking. It’s assumed to require some kind of thinking by analogy so that the person assumes that the other is having the same kind of mental processes guide their behavior that he or she would have in that situation. In other words, it requires knowing one’s own mind first, then imagining others having similar minds. But recent studies have shown that a ChatGPT-3 develops a “theory of mind,” about people’s behavior after being exposed to large samples of language that describe how people behave. This “mindreading” ability emerges without direct teaching. Does that mean that the Chat-GPT-3 has knowledge of its own mind? If so, then such an ability only emerged gradually, because earlier versions of ChatGPT, exposed to smaller samples of language, had less success passing theory of mind tests. In Ezekiel’s Brain I had someone program self-awareness into the AI, but it seems that may not be necessary. It may develop spontaneously or be a property of learning language. This is an area where science may be moving faster than science fiction, or at least my own science fiction.
Just as Arthur C. Clarke anticipated, as more and more is learned about our solar system, science fiction must reach outward, toward the stars. But Proxima Centauri, the closest star to us, is more than four light years away. Getting there would take thousands of years at the speed which current spacecraft can travel. Real solutions to humans getting there involve either generational ships, where travelers live out their lives, then have children who take their place, and so on for several generations, or some form of suspended animation, probably cryogenic. Science fiction writers want their heroes to not only be the same people who started a space journey, but to be able to navigate the ship while taking it. They use two basic approaches: warp drives and hyperspace, the latter also known as stargates, jumpgates, wormholes, subspace, and even unspace, if you read Adrian Tchaikovsky. Following Gene Roddenberry, I opted for warp drive. The most popular and seriously considered version of a warp drive is the Alcubierre Drive, the brainchild of Mexican theoretical physicist, Miguel Alcubierre. Basically, Alcubierre’s idea stemmed from a mathematical solution to overcoming the constraint’s of Einstein’s relativity theory that posited the speed of light as an upper limit for movement without altering space-time. In his model, a spacecraft could exist within a “warp bubble” that contracted space-time in front of it and expanded it behind it so that as it traversed the space in front of it, it needn’t exceed the speed of light because the warp bubble itself was bending space-time. The problem with his invention was that it required an enormous mass to bend space-time, and the expansion of space-time behind the bubble took enormous amounts of energy, negative energy, which is not known to exist except as a mathematical concept. On top of that, the drive had such a gigantic energy demand that it would likely consume anything within millions of miles of it.
Holy USS Enterprise!
In 2021, Erik Lentz wrote a paper showing that a somewhat different version of the warp bubble could work using entirely positive energy. The mass problem still existed, along with other difficulties, such as how to navigate it, but a major impediment to the feasibility of the Alcubierre Drive was eliminated. In Ezekiel’s brain, I use Alcubierrre’s and Lentz’s ideas and put them in the hands of Euclid, an android mathematical savant, who is able to design a workable “Euclidean Drive.”
So much for conscious AIs and warp drives. What about the possibility of finding life on other planets? In Ezekiel’s Brain, the starship Delphi travels to Proxima Centauri, our closest star, and visits Proxima Centauri B, an earth-sized planet that is within the habitable zone and orbits its star once every eleven days. It is also tidally locked, so that the same side faces its star, meaning one side is always hot and bright while the other is cold and dark. All of these known characteristics are taken into account when I describe the planet in the novel.
Proxima b was discovered in 2016, but since then we have launched the James Webb Space Telescope, which orbits our sun a million miles or so from earth. This has enabled us to learn about more distant stars and their planets. In Prime Directive, the sequel to Ezekiel’s Brain, the Delphi visits Trappist-1, a dim dwarf star about 40 light years from earth that is known to have seven orbiting planets, all closer to their weak star than Mercury is to our sun. Observations prior to James Web indicated that at least two of these tidally locked planets were within the habitable zone. In the last year, observations from James Web have suggested that the two innermost planets around Trappist 1 are waterless and atmosphere-less, both rocky and barren and both unlikely to support life. My story happens on the next two planets away from the star, and I make both of them inhabited by human-like species. I’m waiting to see what the next round of James Webb observations has to say about their habitability. Now I know what Arthur C. Clarke felt like when he wrote about a moon station, knowing that by the time the film of 2001 came out in 1968, there might be a real moon-landing to confirm or deny his conjectures. A year after the film, Neil Armstrong walked on the moon. Using real science in science fiction means betting on the chance that the science won’t progress enough to prove you wrong.Want to hear more about putting science into science fiction? Come to Las Vegas on July 15. See below.