What is Artificial Intelligence?
In simple terms, an artificial intelligence (AI) system is a computer that can think like a human. Self-driving cars, online translators, and smartphone assistants like Siri are all examples of AI systems.
AI computers mimic two features of human intelligence:
How is AI different from traditional computing?
To understand this definition, it is helpful to consider how traditional, non-AI computers work. Traditional computer programs follow a set of formal rules (a code), which is written by humans. Accordingly, these programs are strong at performing tasks for which humans can write rules—like solving math equations.
For other tasks, even tasks that humans find easy, it is impossible for us to write explicit rules. For example, a toddler is easily able to distinguish a cat from other animals. But how would we write a code for a computer to identify a cat?
We might start by listing the general characteristics of cats: they have tails, four legs, pointy ears, fur, whiskers, and so forth. These characteristics will be the rules under which our computer program will operate.
Then, when we show our program an image of a cat, it will try and locate the characteristics we listed. However, if a single characteristic is missing from the image—if the cat’s tail is hidden behind its torso, for example—our computer would mistakenly identify the animal as a non-cat.
This would happen because traditional computer programs are inflexible. They have no ability to reason or generalize about the world. Their actions are circumscribed by the instructions that humans write—and when humans fail to write perfectly clear instructions, the program fails at its goal.
AI and Human Intelligence
Humans, on the other hand, are able to make generalizations and think flexibly. When we look at a cat, we do not run through a mental checklist to see if it meets every characteristic of a cat. We instinctively know it is a cat—even if we cannot see its tail—because we understand the general concept of “cat.”
AI systems mimic these characteristics of human intelligence. Using a process called “machine learning,” AI systems can find patterns in huge amounts of data, and use those patterns to make guesses about the world. Machine learning has made AI computers very strong at image recognition. If an AI is shown an image of a cat, it can refer to millions of images of cats that it has seen in the past to determine if there are similarities.
Another parallel between human intelligence and artificial intelligence is adaptability. A self-driving car is able to analyze changes in its environment and respond accordingly, much like a human driver does. For example, if a pedestrian walks onto the street, the car will respond by braking.
For more analysis on how AI works, please see the research packet.