It’s increasingly likely we’re living in the Matrix. Just ask this MIT professor

In the new edition of my book, The Simulation Hypothesis, released in July, I’ve updated my estimate of how likely we are to be in a simulation to approximately 70%, thanks to recent AI developments. This means we are almost certainly inside a virtual reality world like that depicted in The Matrix, the most talked about film of the last year of the twentieth century.

Even young people who weren’t born in 1999 tend to know the basic plot of this blockbuster: Neo (Keanu Reeves) thinks he’s living in the real world, working in a cubicle in a mega software corporation, only to discover, with the help of Morpheus (Laurence Fishburn) and Trinity (Carrie-Anne Moss), that he’s living inside a computer-generated world.

The AI Factor

What makes me so sure that we are living in a simulation? There are multiple reasons explored in the book, including a new way to explain quantum weirdness, the strange nature of time and space, information theory & digital physics, spiritual/religious arguments, and even an information-based way to explain glitches in the matrix.

However, even while discounting these other possible reasons we may in a simulation, the main reason for my new estimate was because of the rapid advance of AI and virtual reality technology, combined with a statistical argument put forth by Oxford philosopher Nick Bostrom in 2003. In the past few years, the rise of generative AI like ChatGPT, Google’s Gemini, and X’s Grok has proceeded rapidly. We now have not just AI which has passed the Turing Test, but we already have rudimentary AI characters living in the virtual world with whom we can interact.

One recent example includes prompt-generated video from Google Veo. Recently, Google has introduced the ability to create realistic-looking videos on demand, complete with virtual actors and landscapes that are completely AI generated, and speak real lines of dialogue, all based on prompts. This has led to prompt theory, a viral phenomenon of AI-generated video of realistic characters exhorting that they were definitely not generated by AI prompts.

Another recent example is the release of AI companions from Grok, which combine LLMs with a virtual avatar, leading to a new level of adoption of the rising wave of AI characters that are already serving as virtual friends, therapists, teachers, and even virtual lovers. The sexy anime girl in particular has led to thousands of memes of obsession with virtual characters. The  graphics fidelity and responsiveness of these characters will improve—imagine the fidelity of the Google Veo videos combined with a virtual friend/boyfriend/girlfriend/assistant, who can pass what I call, the Metaverse or Virtual Turing Test (described in the new book in detail).

The Simulation Point

All of this means we are getting closer than ever to the simulation point, a term I coined a few years ago as a kind of technological singularity. I define this as a theoretical point at which we can create virtual worlds that are indistinguishable from physical reality, and with AI beings that are indistinguishable from biological beings. In short, when we reach the simulation point, we would be capable of building something like the Matrix ourselves, complete with realistic landscapes, avatars and AI characters.

To understand why our progress in reaching this point might increase the likelihood that we are already in a simulation, we can build on the simulation argument that Bostrom’s proposed in his 2003 paper, “Are You Living in a Computer Simulation?” Bostrom surmised that for a technological civilization like ours, there were only three possibilities when it came to building highly realistic simulations of their past (which he called ancestor simulations). Each of these simulations would have realistic simulated minds, holding all of the information and computing power a biological brain might hold. We can think of having the capability of building these simulations as approximately similar to my definition of the simulation point.

The first two possibilities, which can be combined for practical purposes, were that no civilization ever reaches the simulation point (i.e. by destroying themselves or because it isn’t possible to create simulations), or that all such civilizations who reached this point decided not to build such sophisticated simulations.

The term “simulation hypothesis” was originally meant by Bostrom to refer to the third possibility, which was that “we are almost certainly living in a computer simulation.” The logic underlying this third scenario was that any such advanced civilization would be able to create entirely new simulated worlds with the click of a button, each of which could have billions (or trillions) of simulated beings indistinguishable from biological beings. Thus, the number of simulated beings would vastly outnumber the tally of biological beings. Statistically, then, if you couldn’t tell the difference, then you were (much) more likely to be a simulated being than a real, biological one.

Bostrom himself initially declined to put a percentage on this third option compared to the other two, saying only that it was as one of three possibilities, implying a likelihood of 33.33 % (and later changed his odds for the third possibility to be around 20%). Elon Musk used a variation of Bostrom’s logic in 2016, when he said the chances of us being in base reality (i.e. not in a simulation) were one in billions. He was implying that there might be billions of simulated worlds, but only one physical world. Thus statistically, we are by far highly likely (99.99%+) in a simulated world.

Others have weighed in on the issue, using variations of the argument, including Neil deGrasse Tyson, who put the percentage likelihood at 50%. Columbia scientist David Kipping, in a paper using Bayesian logic and Bostrom’s argument, came up with a similar figure, of slightly less than 50-50.

Musk was relying on the improvement in video game technology and projecting it forward. This is what I do in detail in my book where I lay out the 10 stages of getting to the simulation point, including virtual reality (VR), augmented reality (AR), BCIs (Brain Computer Interfaces), AI, and more. It was the progress in these areas over the past few years that gives me the conviction that we are getting closer to the simulation point than ever before.

The Equation

In my new book, I argue that the percentage likelihood we are in a simulation is based almost entirely on whether we can reach the simulation point. If we can never reach this point, then the chances are basically zero that we are in a sim that was already developed by anyone else. If we can reach this point, then the chances of being in a simulation simply boil down to how far from this theoretically point we are, minus some uncertainty factor.

If we have already reached that point, then we can be 99% confident about being in a simulation. Even if we haven’t reached the simulation point (we haven’t, at least not yet), then the likelihood of the simulation hypothesis, P_{sim ,} basically simplifies down to  P_simpoint, the confidence level we have that we can reach this point, minus some small extra uncertainty factor (p_u).

P_{sim  »} P_simpoint – p_u

If we are 100% confident we can reach the simulation point, and the small factor p_u is 1, then the likelihood of being in a simulation jumps up to 99%. Why? Per the earlier argument, if we can reach this point, then it is very likely that another civilization has already reached this point, and that we are inside one of their (many) simulations. p_u is likely to be small because we have already built uncertainty into our P_simpoint for any value less than 100%.

So, in the end, it doesn’t matter when we reach this point, it’s a matter of capabilities. And the more we develop our AI, video game, and virtual reality technology, the more likely it is that at some point soon, we will be able to reach the simulation point.

Are we there yet?

So how close are we? In the new book, I go through each of the 10 stages and estimate that we are more than two-thirds of the way there, and I am fairly certain that we will be able to get there eventually. This means that today’s AI developments have convinced me we are at least 67% likely to be able to reach the simulation point and possibly more than 70%.

If I add in factors from digital and quantum physics detailed in the book, and if we take the “trip reports” of mystics of old and today’s near-death experiencers and psychonauts (who expand their awareness using DMT, for example) at face value, we can be even more confident that our physical reality is not the ultimate reality. Those who report such trips are like Plato’s philosopher who not only broke his chains, but also left Plato’s allegorical cave. If you read Plato’s full allegory, it ends with the philosopher returning to the cave to describe what he had seen in the world outside to the other residents, who didn’t believe him and were content to continue watching shadows on the wall. Because most scientists are loath to accept these reports and are likely to dismiss this evidence, I won’t include them in my own percentage estimation, though as I explain in the book, this brings my confidence level that we are in a virtual, rather than a physical reality even higher.

Which brings us back to the inescapable realization that if we will eventually be able to create something like the Matrix, someone has likely already done it. While we can debate what is outside our cave, it’s our own rapid progress with AI that makes it more likely than ever that we are already inside something virtual like the Matrix.