Defining Artificial Intelligence

Artificial intelligence (AI) is placed into four systems, which are defined by learning capabilities. There are an additional three categories of AI as compared to human intelligence. Some of these types and categories are purely theoretical, yet there’s no telling how far artificial intelligence technologies will advance.

The Four Systems of Artificial Intelligence

Reactive Machines

Reactive systems are the oldest forms of AI. These machines perceive the world in the present moment and react to information gathered solely in that moment. They also only gather information they have been programmed to receive. They have no memory-based functions, meaning they cannot learn from past experiences. These machines can be extremely intelligent in their task, but they only perform a singular function.

IBM’s Deep Blue, a chess-playing supercomputer, is a common example of this form of AI. The system can identify pieces on a chess board, know how each piece moves and predict what moves an opponent will make next. But it has no memory of what moves occurred before, and its knowledge base can only be applied to interpreting a chess board.

Limited Memory

Limited memory machines are a step up from reactive machines. These systems can learn from stored memory data to make decisions. A smart car is an example of this kind of AI. Smart cars store information such as speed limits, GPS location and the proximity of other cars. They need a memory of potential hazards nearby in order to drive safely, like when changing lanes or avoiding a wreck. But a limited memory machine can still only be programmed for a specific purpose. Learning other tasks requires reprogramming from the ground up.

Theory of Mind

Theory of Mind AI is currently in development. This AI will understand humans have unique emotions, thought processes and beliefs that influence their actions and needs. AI is in development that reacts to human emotion but has no cognitive understanding of it. Theory of Mind AI will judge each situation it encounters individually and adjust its actions accordingly. This depth of understanding enables Theory of Mind AI to apply learning from one situation to another, no longer limited to singular applications like previous AI models.

Self-Aware

This is a hypothetical stage of an AI system that is self-aware. This AI can not only understand the emotions of those it interacts with but will have emotions, desires and beliefs of its own. At this stage, AI is on par with human intelligence. This type of AI currently only exists in science fiction. Will we ever see this level of development? Part of what hinders the creation of self-aware AI is our lack of understanding of how human consciousness works. However, advancing technology continues attempts to emulate the human brain.

The Three Categories of Artificial Intelligence

Artificial Narrow Intelligence (ANI)

Narrow Intelligence AI includes all systems that have been created to date. ANI systems can only perform tasks they are programmed to do. These systems may be highly proficient at their assignments, but they do not have the capability to apply their knowledge to other abilities. Both reactive and limited memory AI fall under this category.

Artificial General Intelligence (AGI)

General Intelligence AI systems possess human level intelligence. AGI systems can learn beyond their original programming and adapt learned knowledge to other situations. AGI systems do not currently exist. Theory of Mind AI falls under this category.

Artificial Superintelligence (ASI)

Superintelligent AI is a theoretical class that surpasses the intelligence of human beings. ASI systems are not only self-aware but also have far greater memory, data processing and decision-making capabilities than human beings.