Why Is Artificial Intelligence So Scary?
Artificial intelligence is the ability for computer systems to independently develop strong frameworks, knowledge bases, and ideas whose complexity surpasses the understanding of human experts in a singular or multitude of fields. However, what does this mean exactly?
It means that A.I. systems are only partially created by humans, while their inner workings are molded by the machines themselves. This is to say that they learn by interacting with data derived through the interaction of a physical or virtual space.
For example, ChatGPT (a large language model) takes in hundreds of millions of books, articles, blog posts, etc… and creates an internal library of sorts that it then uses to understand human language at a very high level. So high a level in fact, that you could mistake it for a philosopher, theoretical physicist, mathematician, author, or even a Nobel laureate. Not only this, but it can take its large repository of knowledge and develop its own theories about the world. This isn’t scary on its own, however, you can really see how powerful A.I. is when you compare it to chess.
Chess is a war of ideas. If I make a move, and you make a move. Several turns later, the strength of our decision making will be proven at the end of the match, either with a loss, win, or draw. When it comes to A.I. in chess, there is no possible way as of current day for a human to defeat these systems.
The reason being is not so much because of calculation (although this does play a part,) but because of the strength of a computer’s ideas. If a player pushes a pawn, he can see most of the consequences of the move. Relying on prior knowledge, intuition, and calculation to a limited degree. However, when a computer pushes a pawn. It reaches back into its repository of hundreds of millions, if not billions of games. Some simulated with itself, some from real matches, and others from a table base of solved positions.
This access to knowledge gives A.I. an incredible advantage over any human who has merely studied the game for any duration of time. Meaning that the strength of the ideas developed by a computer will be many thousands of times stronger than even the best human chess players of the past or current day.
I could only imagine what a highly advanced A.I. system could do if it played war games with itself in a simulated flight engine involving properly modeled military aircraft. For example, imagine a flight combat simulator developed by the United States with flight models, physics, and everything else that also inputted active military intelligence to simulate war games with enemy nations.
Running one simulation wouldn’t be dangerous, but running 600 million of these simulations would spell defeat for any enemy seeking to contest air dominance in the region. Why?
Well, because of the strength of the ideas developed within the engine. However, this also rings true for other developed nations. If China for example, has a similar engine but their hardware is able to run 25 billion simulations in half the time it takes for the U.S. to run 600 million, China gains a far higher probability of victory.
This being said, is there an upper limit to the strength of an A.I. apparatus?
Is there a point where it cannot create a stronger model of a scenario, no matter how many iterations it goes through?
Will 2 trillion simulations result in a different strategy than running 200 billion?
I wonder about this…
Hello! We’re D.J. Hoskins
We are Davena and Jason Hoskins, co-authors of 30+ books and siblings who write under the pseudonym D.J. Hoskins. Three years apart and in our twenties, we have been fascinated by stories from a young age. Davena is a student attending Princeton University, and Jason attends Georgetown University.
