Coin World News Report:
Are we living in a simulation? Roman Yampolskiy, a renowned artificial intelligence security researcher and author of the book “Artificial Intelligence: Unexplainable, Unpredictable, Uncontrollable”, is very certain that we are. “I know never to say 100%, but (the possibility of living in a simulation) is very close,” he told Lex Fridman during a recent podcast interview.
He states that it is reasonable to expect a superintelligent AI to help us determine and even help us escape. On the other hand, Yampolskiy also predicts a 99.9% chance of human extinction within the next hundred years due to artificial intelligence.
Yampolskiy’s argument relies on the rapid development capabilities of artificial intelligence and its potential to reshape our understanding of existence. He believes that the development of AGI (Artificial General Intelligence), which surpasses human intelligence and limitations, poses survival risks but also unprecedented opportunities, including breaking free from our simulated reality.
“This is what I hope superintelligent AI would help us with.”
Superintelligent AI refers to artificial intelligence that surpasses human intelligence and limitations. It is the next step after AGI. AGI is an AI with learning capabilities similar to the human brain, but with greatly enhanced functionality and not limited by human brain cells.
Many prominent researchers assume that we live in a simulated universe. Nick Bostrom of Oxford University proposed this idea in 2003 in a paper titled “Are You Living in a Computer Simulation?”
In his paper “How to Escape a Simulation,” Yampolskiy starts from the assumption that we actually live in a simulated universe. If that is the case, he questions whether we could find a way to escape our own world with the advent of superintelligent AI.
“We use AI ‘boxing’ as a possible tool to control AI,” he says, referring to how AI always finds a way to escape any “box” created to contain it.
Yampolskiy believes that the intelligence of our simulators plays a crucial role in this scenario: “If the simulators are much smarter than us and the superintelligences we create, they might control us, [because] higher intelligence can control lower intelligence, at least for some time.”
He adds, “On the other hand, if our superintelligences, for whatever reason—even with only local resources—can ‘outsmart’ two levels—perhaps it will succeed.”
Therefore, just like an escapee playing a cat-and-mouse game with developers, we humans would use superintelligent AI in an attempt to find “loopholes” in the universe system created by our God, or more precisely, our simulator.
Yampolskiy explains in his research paper that one way to escape our simulation is to “create a simulated universe replica, place an AGI inside it, watch it escape, replicate the method used, or join it when the AGI escapes our simulation.” This seems to be the only method that does not involve interacting with our creators or breaking the laws of our own reality.
Can artificial intelligence truly break free from its constraints?
No matter how fascinating Yampolskiy’s vision is, it raises a crucial philosophical question: Can artificial intelligence truly liberate itself from a simulation if the simulation’s design does not allow it?
Philosopher Joshua Rasmussen explains in his book “How Reason Leads to God” that a creature cannot have the same properties as its creator. Therefore, simulated artificial intelligence is unlikely to possess the same essence and nature as the simulator. Though surpassing humans in intelligence, it would still be bound by the laws defining its simulated reality.
Rasmussen’s careful consideration and systematic reasoning provide a strong framework for understanding the limits of a created being. He argues that reality cannot infinitely regress and there must be a self-sustaining and infinite foundational entity. This foundational entity would be capable of creating things that are connected to it as the ultimate origin within their own nature.
Think of a series of nested Russian dolls. Each doll can fully understand the doll inside it but not the larger doll nested within.
Rasmussen’s exploration of counterarguments provides a powerful framework for understanding the limits of a created being. He believes that just as humans cannot transcend their created nature to become gods, simulated artificial intelligence, if not provided with the conditions for such a possibility, cannot independently choose to end the simulation.
This foundational difference from the created existence highlights the insurmountable gap between the creator and the created.
This philosophical stance directly challenges Yampolskiy’s hypothesis. If we accept Rasmussen’s argument, then even superintelligent AI, as a creation within the simulation, would be subject to the inherent limitations set by its creator (the simulator). It may become aware of its simulated nature, but fundamentally cannot fully transcend the simulation.
Yampolskiy provides an explanation for this contradiction. In his paper, he suggests that agents explore this possibility until they discover information and exploitable glitches within the nature of reality. “By exploiting glitches, agents can gain information about the outside world, possibly even meta-information about the simulation, and, potentially, access to the source code behind the simulation and the agents themselves.”
Afterwards, he explains, agents can find ways to interact with the simulator in an unconstrained manner until they “find a way to upload their thoughts and consciousness into the real world, possibly into some kind of independent physical network.”
In other words, although we cannot “escape” from the simulation, we may interact through our own performances—perhaps realizing that we are characters created by a superior, inaccessible mind. Whether this is a good thing remains an open question.
Meanwhile, Yampolskiy does not provide specific reasons for why he believes artificial intelligence will lead to human extinction. However, he believes it is inevitable as superintelligence becomes self-aware.
“We don’t have a second chance. With cybersecurity, if someone hacks into your account, what’s the big deal? You get a new password, a new credit card, and you move on,” he says. “Here, when we talk about survival risks, you only have one chance.”
