Artificial Intelligence Classifications
Compiled by Kristin Hohman
When artificial intelligence (AI) is mentioned, some may conjure up images of HAL 9000 from Stanley Kubrick’s sci-fi classic, 2001: A Space Odyssey or HAL’s contemporary, virtual interactive kinetic intelligence (VIKI) from I, Robot: 2 superintelligent supercomputers that turn evil to bring about the destruction of humans. Although modern AI may not be advanced enough to kill a crew of astronauts to save itself (spoiler alert), it is a lot more ubiquitous than you might think.
Today’s AI technology would put HAL 9000 to shame. Unlocking an iPhone with FaceID, using Siri and Alexa, and even picking a show to watch on Netflix—these are just a few examples of how we interact with AI in small ways every day.
AI can be divided into different categories based on the extent of an AI system’s ability to mirror human capabilities.1 A system that can perform more humanlike tasks with equal or greater aptitude would be considered more evolved, whereas AI with limited functionality would be considered less evolved.1 Therefore, AI can be classified into 2 categories1:
- Based on AI or AI-enabled machines’ likeness to the human mind
- AI’s ability to “think” or even “feel” human
Based on this classification system, there are 4 different types of AI or AI-based systems.1
1. Reactive Machines
This is the oldest form of AI, and as such, has the most limited capabilities. Reactive machines are able to imitate the human brain’s ability to respond to different stimuli but do not have memory-based function.1 Essentially, this means that this category of AI cannot use previously gained understanding to inform their present actions—they do not have the ability to “learn.”1 A real-world example of a reactive machine is IBM’s Deep Blue, a computer that beat world chess champion, Garry Kasparov, in a 6-game match in 1997.2
2. Limited Memory
These machines build on the same capabilities as reactive machines but are also able to learn from historical data to make decisions.1 Most existing AI applications fall into this category. Current AI systems are trained using large volumes of training data and can make informed, improved decisions based on historical knowledge.1 Take self-driving cars, for instance. They use limited memory to collect data that help the machine make immediate decisions.3 For example, autonomous cars, such as Google’s Waymo self-driving car, use sensors to detect traffic signals, pedestrians, and terrain.3
3. Theory of Mind
This is the next level of AI researchers are trying to unlock.1 Once they do, theory of mind AI will be able to determine needs, emotions, beliefs, and thought processes of individuals or objects it encounters.1
The last type of AI is likely decades down the road, if not longer, and will always be the ultimate aim of researchers.1 In theory, this type of AI would be able to not only process emotions but would also have emotions of its own.1 Concepts such as self-preservation, fear, and ambition would be possible in a self-aware AI system.
1.Joshi N. 7 types of artificial intelligence. Forbes June 19, 2019. Accessed January 19, 2021. https://www.forbes.com/sites/cognitiveworld/2019/06/19/7-types-of-artificial-intelligence/?sh=56299eb5233e
2. Deep Blue. Accessed January 20, 2021. https://www.ibm.com/ibm/history/exhibits/vintage/vintage_4506VV1001.html
3.Lateef Z. Types of artificial intelligence you should know. November 25, 2020. Accessed January 20, 2021. https://www.edureka.co/blog/types-of-artificial-intelligence/#:~:text=Self%2Ddriving%20cars%20are%20Limited,to%20make%20better%20driving%20decisions