Smart & Connected Life Smart Home The Four Types of Artificial Intelligence What to know about the different levels of AI by Sam Costello Writer Sam Costello has been writing about tech since 2000. His writing has appeared in publications such as CNN.com, PC World, InfoWord, and many others. our editorial process Facebook Twitter Sam Costello Updated on February 23, 2021 Tweet Share Email Smart Home Your Best Year Ever: College Tech Tips Amazon Appliances & Lighting Google Artificial Intelligence, or AI, is increasingly becoming a part of our everyday lives, even if we don't always know it. Because it's already here and only going to become more critical in the future, it's essential to understand the four types of artificial intelligence, how they're distinct, and which ones are in use today. Overview of Artificial Intelligence Artificial Intelligence is measured in comparison to the functions of the human mind. A true, fully complete AI could replicate—and likely extend—a human brain's functions. It would be self-aware, have consciousness and memories, could learn and make predictions about likely future events—and would do these things on its own, without human intervention or input from programmers. The Four Types of AI: Reactive Machines The first and most basic kind of AI is Reactive Machines. These machines have intelligence that exists almost entirely in the present moment (eat your hearts out, meditators!). Programmers build Reactive Machines around complex sets of rules. In addition to those rules, they may include neural networks that allow them to learn and adapt in the moment. Crucially, though, this type of AI can't form memories or act based on past experiences. Because they can't create memories, they can't take past occurrences, even those that happened to them, and use them to make new decisions. In this way, they're restricted to the scenarios and information covered in their sets of rules. Because of this limitation, one can trick Reactive Machines into being defeated by forcing them to operate outside their limited set of rules. Reactive Machines also don't have a concept of the future. They can use their rules to make extrapolations about what might come next, but only within the boundaries established by their rule sets. Because they can't change their future actions based on past occurrences, Reactive Machines can't learn. Examples: IBM Deep Blue (plays chess), IBM Watson (won at Jeopardy), Google AlphaGo (plays Go), recommendation engines like those found in streaming video and audio services Learn more about Reactive Machines AI The Four Types of AI: Limited Memory This more-advanced AI type has the abilities of reactive machines but adds a concept of the past. While Limited Memory AIs aren't forming memories, they are aware of a recent past and can use the data captured at that time to influence their decisions. Self-driving cars are the best example here. They continuously monitor the conditions around them—what other vehicles are doing, where objects are, how pedestrians are moving, etc.—and holding that information in a temporary state to influence their actions. As you might imagine, Limited Memory AIs process tremendous amounts of data and make decisions very quickly. This AI level is called "Limited Memory" because these past experiences aren't stored permanently, and it can't use them for all future learning. Instead, these past experiences are saved for a short time while they're practical and then discarded. Examples: Self-driving cars Get an in-depth view of Limited Memory AI The Four Types of AI: Theory of Mind With Theory of Mind AI, we get closer to the science-fiction ideal of artificial intelligence. This level of AI includes all of the aspects of Reactive Machines and Limited Memory. But it adds a crucial and complex bit of understanding: that living objects (people, animals, etc.) around them have a thing called a "mind." Once an AI understands that other creatures have minds, it can then understand that it should learn and adjust its decisions based on those minds. It then understands minds generate thoughts and emotions (even if it doesn't truly understand what those things are yet) and that thoughts and emotions will influence behavior. The AI has to use those factors in its decision-making framework. Examples: There are no known examples of Theory of Mind AI in existence (though some robots that attempt to understand and simulate point a direction for this type) Learn more about Theory of Mind AI The Four Types of AI: Self Aware Self-Aware AI is the final and most complex, and mature level of AI. It's a genuinely self-aware intelligence. Any AI from science fiction falls into this category. It's an entity with true consciousness. It's aware of its existence and its internal states (and potentially emotions), can form memories of the past, and make predictions. It's aware of other consciousnesses and can take them into account when making decisions. Crucially, it can learn and become more intelligent based on its experiences. A Self-Aware AI would require extremely flexible programming logic, an ability to update its logic on its own, and a tolerance for inconsistency since human behavior isn't always neatly predictable or rigidly patterned. Examples: Self Aware AI doesn't exist currently and is likely many years away from being achieved Learn more about Self Aware AI The Four Types of AI Compared The four key types of AI, and their key characteristics, are listed below for easy reference and comparison. AI Type Follows Rules Can Learn Has Memories Knows Other Minds Exist Aware of Emotion True AI Reactive Machines ✓ Limited Memory ✓ ✓ limited Theory of Mind ✓ ✓ ✓ ✓ ✓ Self Aware ✓ ✓ ✓ ✓ ✓ ✓ Should We Worry About AI? Anyone passing familiar with science fiction about AIs knows that things often go wrong for humans once a machine achieves consciousness (hello, Skynet). So, should we be worried about AI? There's no single answer to this question, but it's a good idea to be thoughtful and careful about how we go about creating and using fully self-aware AI. There's the doomsday scenario of AIs replacing humans—whether that means taking human jobs and leaving people without work or income or the darker Terminator-style storyline. There are also ethical concerns: Is it acceptable to create a consciousness that can think and feel and then force it to do our bidding? Some ethicists study AI and write about these questions. As AI becomes more advanced and more widespread, we'll need to make sure we listen to them, and our laws and governments adapt to the unique challenges—and possibilities—created by AI. Was this page helpful? Thanks for letting us know! Get the Latest Tech News Delivered Every Day Email Address Sign up There was an error. Please try again. You're in! Thanks for signing up. There was an error. Please try again. Thank you for signing up. Tell us why! Other Not enough details Hard to understand Submit More from Lifewire How AI Can Manipulate Your Choices The 8 Best Driverless Car Manufacturers of 2021 AI Can Help Pick Your Best Photos The 5 Best Online Flashcards of 2021 How AI Could Change Home Buying Mobile Technology: AI in Phones How AI Changes Our Perspective on Fine Art Argo AI: The Ford-backed Driverless Car The 6 Best 75-Inch TVs of 2021 What Are Autonomous Cars? GM Cruise Driverless Cars: What They Are 10 of the Most Popular Science and Education YouTube Channels The 10 Best Free Online Classes for Adults in 2021 5 Computer Networking Trends for 2021 and Beyond What is a Deep Fake? What Is Google DeepMind?