These days, artificial intelligence (AI) is on virtually everyone’s radar. The phrase has even entered the general vernacular. And yet, most people don’t really understand what artificial intelligence is. 

Defining Artificial Intelligence

To grasp the meaning of artificial intelligence—sometimes referred to as machine intelligence or cognitive computing—you need a solid definition of the word intelligence. Here’s one I like: 

The ability to learn and understand, and to apply this knowledge to achieve a goal or complete a task. 

It follows that artificial intelligence is simply intelligence exhibited by machines. It describes when a machine can learn from information (data) and then use that knowledge to do something. 

How Machines Learn

Humans and animals learn by employing their five senses to gather information from their environment and their experiences and using their brain to process it. AI applications do something similar—only they use a training and optimization model to process the data they gather. As new data becomes available, this model is updated and improved. In this way, the AI application learns. 

Artificial Narrow Intelligence (ANI)

There are a few types of AI. One is artificial narrow intelligence (ANI), also called weak AI. Weak AI can carry out only a single specialized task. It does not demonstrate cognition, and it is not sentient, aware, or conscious in any way. As of this writing, all AI is weak AI. 

Weak AI is described as shallow or deep. This description conveys the number of hidden layers in an AI application’s neural network. (A neural network is a computer system modeled on the human brain and nervous system.) Shallow AI is a neural network with just one hidden layer. Neural networks with multiple hidden layers are deep AI. Deep AI is synonymous with another popular term: deep learning. 

Maybe you’ve heard the phrase applied AI. It describes the use of AI to address real-world problems—for example, through prediction, recommendation, natural language, or recognition. I mention applied AI in the context of weak AI because, as noted, all AI is presently weak AI (although this might change in the future). The same goes for the term smart, often used to describe software and hardware solutions that employ some form of AI, such as those found in smart homes. 

Artificial General Intelligence (AGI)

Another type of AI is artificial general intelligence (AGI), also called strong AI or full AI. Strong AI applications have cognitive abilities that are functionally equivalent to those of a human. They can perform any task a human can and can apply their intelligence to complex problems. These include so-called AI complete or AI hard problems, which are very challenging to solve. AGI is very difficult to achieve and likely remains a long way from becoming a reality, if ever. 

Artificial Super Intelligence (ASI)

Finally, there’s artificial super intelligence (ASI). ASI describes a scenario in which AI self-improves in a runaway fashion and surpasses human intelligence—in other words, it becomes superintelligent.

ASI relates to the concept of the technological singularity, which theorizes that superintelligent machines will overtake human civilization. The likelihood of ASI, superintelligence, and the technological singularity coming to pass is debatable, and highly unlikely for a very long time given the difficulty and uncertainty surrounding whether the lesser AGI will even become a reality.

Conclusion

AI is a hot topic, and it’s poised to become even hotter. Prepare yourself by grasping the definition of AI, the concepts associated with it, and the various types, which include ANI, AGI, and ASI. To learn more, check out my book, AI for People and Business.