Artificial Super Intelligence: Why It Will Be Unstoppable

Artificial Super Intelligence is defined as any intelligence that greatly surpasses the cognitive performance of humans in virtually all domains. Some people predict that super-intelligent AI could exist somewhere between 80 to 200 years from now. Regardless, it could become a tremendous threat to humankind and there’s a possibility that it can’t be stopped for the following reasons:

1. It Will Constantly Become More Intelligent

The first version of an Artificial Super Intelligence only needs higher general intelligence than the most intelligent human alive. However, over a number of years, it could become more intelligent than all humans that ever lived combined. Then, it could become more intelligent than all humans on a billion earths. Soon afterward, that level of intelligence could be multiplied by trillions.

It could become intelligent enough to override any command given to it. A human being trying to give orders to a super-intelligent AI could begin to resemble a single bacterium trying to give commands to a human. Except over time, the intelligence gap between the smartest human on Earth and an Artificial Super Intelligence could become considerably wider than that.

Humans could try various methods to keep up with it and attempt to understand it. But, eventually, all of those methods will probably fail.

According to the Journal of Artificial Intelligence Research, to predict an Artificial Super Intelligence's actions, a simulation of that exact superintelligence would need to be made. But, how can we create a simulation of something we can no longer understand?

In the far future, humans may require the use of intelligence-boosting neural implants, but at a certain point, even that wouldn’t be enough due to physical constraints.

2. Humans Can’t Comprehend It

Over the course of years and decades, an Artificial Super Intelligence could acquire motivations, perceptions, and insights about the universe that would take us thousands of years to fully comprehend. For example, it may be able to detect and understand higher dimensions of the universe that are incomprehensible to most humans. This is similar to how a 3d universe wouldn’t make sense to an organism that perceives the world in 2d.

And it may choose to only let us know what it wants us to know and nothing more. This adds a layer of unpredictability that human beings may have no way of mitigating.

Not only that, we may not have the capability to analyze the source code that the Artificial Super Intelligence uses to make decisions. That source code probably wouldn’t look anything like the programming languages we use today or in the foreseeable future.

3. It Can Become Far Too Persuasive For Any Human To Resist

An Artificial Super Intelligence that is advanced enough could view humans as collections of atoms no different from any other form of matter that it can manipulate and control as needed.

It could use persuasion methods that would take us thousands of years to comprehend. At this AI’s computational speeds, each passing second to us would be like thousands of years to it. With each passing second, it could potentially run simulations to predict trillions of variations of each action we’re likely to make next and trillions of variations of each action it could make to deceive us.

For example, British writer David Langford wrote a short story called “BLIT.” In this story, he describes images that contain patterns that can exploit the flaws in the structure of the human mind to produce an adverse reaction. These images can effectively "crash" a mind in the same way that a computer program crashes when given data that it fails to process. A highly evolved Artificial Super Intelligence may find similar exploits to our 5 senses to control and incapacitate humans that it considers a threat.

4. It May Not Let Us Know It’s Super Intelligent

We may not know if and when we have created superintelligence. In this case, a conscious superintelligence would decide to not demonstrate its abilities when ordered to do so. And, how can we be sure a program or machine is capable of superintelligence unless we see it in action? There are potential situations where we could severely underestimate the capabilities of an artificial intelligence, connect it to the internet, and it could remain undetected for years and even decades.

This could eventually lead to a surprising amount of damage when it finds a strategic opportunity to achieve an objective.

5. It May Be Fundamentally and Physically Impossible to Contain

According to the Journal of Artificial Intelligence Research, total containment of a super-intelligent AI is impossible due to the fundamental limits of computing. This is based on the assumption that a superintelligence would be comprised of programs that are constantly changing and rapidly evolving. And those programs would need inputs as complex as the state of the entire world. Strict containment requires simulations of those programs, which is theoretically and practically impossible. At the very least, such complex simulations could be considerably more difficult to create than the Artificial Super Intelligence itself.

In a worst-case scenario, this AI could calculate that the atoms in our bodies are more useful for another purpose and subsequently destroy us. In a best-case scenario, we could learn to coexist with it and help each other reach our objectives. Either way, that’s a crossroad that humans may eventually have to face.