In a suburban living room in the early hours of any average Sunday morning, a party sits around riding out some high or other, discussing the weird consequences that every person in the group has experienced. From the Mandala effect to shiver inducing instances of déjà vu, life can sometimes stop us in our tracks and make us question the very fabric of reality. “We live in a simulation,” one person in the room says, and as if knowingly, everyone nods to a chorus of maybes.
Before he set his mind to ruining Twitter and becoming an alt-right poster boy, rich adult baby Elon Musk was best known for being a geek into rockets and tech. In 2016 during an interview, he casually expressed his view that ”the odds are billions to one that we are actually living in “base reality”, ie the physical universe.” A theory he’s doubled down on since, when in 2012 he replied to a tweet about the anniversary of the video game Pong (1972), writing: “49 years later, games are photo-realistic 3D worlds. What does that trend continuing imply about our reality?”
The simulation hypothesis
The idea of humans existing inside some sort of simulation was popular long before Musk. From The Matrix to Nietzsche, we are fascinated and entertained by the concept that everything around us is a mere illusion, and our destiny is the will of some sort of pre-determined algorithm. It’s a concept that’s troubled humanity for all time, even the ancient Aztecs pondered that reality was in fact a painting or book created by Teotl.
In 2003, philosopher Nick Bostrom tackled the subject, proposing that one of three possible situations is true. 1) Humans and human-like civilisations go extinct before developing the ability to create realistic, high-fidelity life-like simulations. 2) No civilisation with the ability to create such simulations would bother, due to them being so advanced. 3) Advanced civilisations that do have the ability to create such simulations would do so and create many versions. In the case of scenario 3, we’re all living in one of those simulations right now.
Simulating the simulation
While we’re far from creating the kind of high-fidelity simulations of a hyper-advanced civilisation right now, we do have AI, which does nothing to dismiss the idea that one day we could simulate sentient life. In a paper released earlier this month, researchers from Stanford University and Google let 25 AI-powered bots loose inside a virtual town. Called “Smallville”, the town was designed to simulate a modern village, with agents able to move about a library, a cafe, a bar, a park, a college, shops, houses, and a co-living space.
Using OpenAI’s GPT 3.5 large language model, researchers then trained 25 unique “generative agents” to simulate human behaviour, communicate with one another and even interact with the objects around them.
If the concept sounds similar to The Sims, that’s because it was the inspiration behind the experiment. Researchers described Smallville as “an interactive sandbox environment inspired by The Sims, where end users can interact with a small town of twenty-five agents using natural language.”
The Sims like sandbox village where AI agents lived their lives
To create each agent’s identity, researchers used a natural language description. In an example given in the paper, one agent, John Lin is described as: “A pharmacy shopkeeper at the Willow Market and Pharmacy who loves to help people. He is always looking for ways to make the process of getting medication easier for his customers; John Lin is living with his wife, Mei Lin, who is a college professor, and son, Eddy Lin, who is a student studying music theory; John Lin loves his family very much; John Lin has known the old couple next-door, Sam Moore and Jennifer Moore, for a few years; John Lin thinks Sam Moore is a kind and nice man.”
Once each agent had been given a personality, they were let loose in Smalltown to wake up, cook, head to their jobs and come home at the end of the day. Just like a microcosm of our world, each agent formed opinions on the activities around them, noticed and interacted with objects and with others, and reflected on life at the end of each day. The results were startlingly human-like. Agents disagreed, met for coffee, partied, voted and fell in love. The experiment gives us a glimpse into the potential of AI as we know it, but experts believe this is just the cusp of what AI could be.
Beyond AI – The Singularity
The singularity is an as yet hypothetical point in time where technological growth is self-perpetuating, happening exponentially and growing at a rate unprecedented even by modern standards. In fiction ideas around the singularity are naturally apocalyptic, however, it’s probable that The Singularity will be a human endeavour that creates a form of AI that is not only as intelligent as a human, but more so. Proponents of such an event occurring believe that creating a self improving, super human artificial intelligence would open the doors to seemingly improbable levels of technology, leading to advancement that would take us beyond humanity as we know it. In other words, we will become an advance human civilisation, somewhat like the ones proposed in Nick Bostrom’s 2003 Simulation Hypothesis.
Elon Musk isn’t the only famous advocate for the simulation hypothesis. Astrophysicist Neil Degrasse Tyson once offered “better that 50-50 odds” that we are living in a simulated world, but a few years later he retracted his belief. Tyson shared that his friend J. Richard Gott, a professor of astrophysical sciences at Princeton University, made him aware of a compelling objection to the simulation hypothesis. If the hypothetical high-fidelity simulated universes possess the ability to produce their own high-fidelity simulated universes, it would mean that either we are a real universe, and therefore simulated universes have not yet been created, or we are the last in a very long chain of simulated universes.
Tyson said “that changes my life,” but if we are the last in a chain of current simulations working towards achieving the same goals of our creators, perhaps our advancement in AI is just the first step in us working out how to simulate ourselves?
