If We're Living In A Simulation, A Computer Scientist Has A Plan To Escape

If we're in a simulation, there are ways we can escape.

James Felton

James Felton

James Felton

James Felton

Senior Staff Writer

James is a published author with four pop-history and science books to his name. He specializes in history, strange science, and anything out of the ordinary.

Senior Staff Writer

A big red sign saying "exit".
Would you want to escape? Image credit: David Schliepp/

If we were in a hyper-realistic simulation, à la The Matrix, would it be possible to escape? Computer scientist Roman Yampolskiy has outlined in a new paper how we might escape, and what that would even look like.

Simulation Theory, in its most basic form, goes like this: if humans (or another species, for cuteness feel free to imagine it's puppies) continue to advance for hundreds, thousands, or even millions of years, it's a pretty safe bet that we will have a lot of computational power at our finger/paw tips. If we were to expand out into the galaxy (or even further) we may harness the power of stars, or possibly even black holes.


With all this energy and computational power, it's likely that at some point our descendants will be curious enough to run "ancestor simulations", using just a tiny fraction of the computing power available to us. 

Ancestor simulations, as put forward by Swedish philosopher and Oxford University Professor Nick Bostrom in his 2003 paper "Are you living in a computer simulation?", is the idea that future generations might have the computing power to run simulations on our forebears, and imbue these simulations with a sort of artificial consciousness. If this has already happened, it would mean the vast majority of people are simulations by the advanced descendants of the original humankind, and if that's the case, it's more rational to assume you are one of the simulations rather than one of the original biological humans. 

In his paper, Bostrom proposes three possible scenarios:

1) The fraction of human-level civilizations that reach a stage where they are able to run these simulations is very close to zero. 


I.e. it's likely we'll get wiped out before we reach a point where we are able to perform such tests.

2) The fraction of posthuman civilizations that are interested in running ancestor simulations is very close to zero. 

In other words, our species has changed so much by that point that we are no longer interested in running simulations, and no curious individuals have access to the power to create them, or else running these simulations is banned.

3) The fraction of all people with our kind of experiences that are living in a simulation is very close to one.


If the other two are false, then we are left with option number three: our species develops the required technology and starts creating an incomprehensible number of ancestor simulations (over time). This would mean that the vast majority of "people" with experience of living on Earth are... inside a simulation, we just don't know it. 


So, say we're in a simulation. How do we get out? In a new paper, Roman Yampolskiy, a computer scientist at the University of Louisville, attempts to answer that very question, and suggests a few ways of busting out of here.

First, he writes, it is best to try and figure out what kind of simulation we are in.

"We can postulate two main types of simulations we could be in; partial-simulation in which a virtual environment is simulated and into which non-simulated agents are immersed, akin to what we call Virtual Reality (VR)," he writes in the paper, "and full-simulation in which both environment and agents (us) are generated."


"A partial-simulation implies that triggering a shutdown may be sufficient to get back to the base reality, while a full-simulation would require a more sophisticated approach." 

One method would be to force our simulators to use ever-increasing computational power until they can no longer ignore it. 

"Perhaps we could send off Von Neumann probes to the far corners of the universe, in a bid to deliberately increase re‐source consumption," Yampolskiy cites programmer Gwern Branwen, "or we could run simulations of our own".

The idea is similar to a "brute force" attack. In this case, we would try and force the simulator to use more and more power, until they wonder why their version of Chrome is loading slowly and take a look at the background processes, or the simulation itself crashes.


The risk, of course, is that we simply get shut off. Say we are in a partial simulation, this means we could get out into whatever world is out there, but if we are simulated as well, that's bad news for anyone who likes not being ctrl-alt-deleted out of existence.

The paper, which is well worth a full read if you are interested in the topic, runs through a number of suggestions from others for escaping the system. They range from trying to attract the attention of the creators through a gigantic monument in binary to let them know that we know, to deliberately creating a time-travel paradox by murdering the time traveler's grandfather, causing the computer to crash as it tries to reconcile the paradox. 

One intriguing idea, gained from an anonymous fiction story posted to the Internet in 2014, is that we "hack" the simulation and escape it using any exploits that may have been left around the universe. Bizarrely, there is a sort of analog to this in the game Super Mario World.

In 2016, YouTuber Seth Bling was able to hack a copy of Super Mario World using only movements in the game, performing a specific series of actions in order to gain the ability to affect the game's code. Through writing instructions by (among other things) blasting fire from Yoshi's mouth, he was able to extend the level's timer, then eventually turn Mario into a version of the game Flappy Bird.


"Since it was possible to write code with precise Mario movements and spin-jumps," Yampolskiy writes, "that implies that if Mario was sufficiently intelligent he could discover and code this hack from within the Super Mario World (assuming Mario’s actions are writing to the same memory locations as actions from the controllers used to generate Mario’s actions)."

He then quotes an ancient magic spell. 

“Take a lion cub and slaughter it with a bronze knife and catch its blood and tear out its heart and put its blood in the midst ... and write the names of … angels in blood upon the skin between its eyes; then wash it out with wine three years old and mix … with the blood.” 

Probably not entirely seriously, he suggests that maybe these are in fact hacks of the universe, though "we don’t have sufficient meta-data which can explain why all magical spells fail to work in practice even if they corresponded to working hacks in our universe".


Our easiest escape, though, would be to attract the attention of an observer of the simulation and convince them to aid our exit into the real world, perhaps by eliciting empathy for our situation.

However, for his actionable plan, Yampolskiy writes that we are now in the first early stage of researching the possible ways to escape. The next step would be to investigate the structure of the universe more (which we are already doing, but for other reasons) and particularly quantum mechanics. 

"As we currently have no capability to read/write simulation’s source code and do not know if our attempts at social engineering attacks have any impact, our best bet is to investigate the structure of our universe at the smallest possible scale in the hopes of detecting exploitable effects," he wrote, adding that quantum mechanics has plenty of weirdness to it, which would "make a lot of sense" if we saw them as glitches or possible exploits.

"Such anomalies, alone or in combinations have been exploited by clever scientists to achieve what looks like simulation hacking at least in theory and often in later experimentation (ex. modifying the past, keeping cats both dead and alive, communicating counterfactually)," he continued. 


"While the quantum phenomena in question are typically limited to the micro scale, simply scaling the effect to the macro world would be sufficient for them to count as exploits in the sense used in this paper."

Of course, if we do escape, there is no telling what the actual world (or the simulation above ours) will be like.

A preprint of the paper is available on ResearchGate.


  • tag
  • simulation,

  • computer science