7 Types of Artificial Intelligence

If you’ve ever used Amazon’s Alexa, Apple’s Face ID or interacted with a chatbot, you’ve interacted with artificial intelligence (AI). 

There are a lot of ongoing AI developments, most of which are divided into different types. These classifications reveal more of a storyline than a taxonomy, one that can tell us how far AI has come, where it’s going and what the future holds. 

These are the seven main types of AI to know, and what we can expect from the technology.

Capability-Based Types of Artificial Intelligence

Based on how they learn and how far they can apply their knowledge, all AI can be broken down into three capability types: narrow AI, artificial general intelligence and artificial superintelligence.

1. Narrow AI

Narrow AI, also known as artificial narrow intelligence (ANI) or weak AI, describes AI tools designed to carry out very specific actions or commands. They are built to serve and excel in one cognitive capability, and cannot independently learn skills beyond their design. All AI systems used today fall under the category of narrow AI.

Narrow AI often utilizes machine learning, natural language processing and neural network algorithms to complete specified tasks. Some examples of narrow AI include self-driving cars and AI virtual assistants.

2. Artificial General Intelligence (AGI)

Artificial general intelligence (AGI), also called general AI or strong AI, refers to a theoretical form of AI that can learn, think and perform a wide range of tasks at a human level. The ultimate goal of AGI is to create machines that are capable of versatile, human-like intelligence, functioning as highly adaptable assistants in everyday life. 

Though still a work in progress, the groundwork of artificial general intelligence could be built from technologies such as supercomputers, quantum hardware and generative AI products like ChatGPT. 

3. Artificial Superintelligence

Artificial superintelligence (ASI), or super AI, is truly the stuff of science fiction. It’s theorized that once AI has reached the general intelligence level, it will soon learn at such a fast rate that its knowledge and capabilities will become stronger than that of even humankind. 

ASI would act as the backbone technology of completely self-aware AI and other individualistic robots. Its concept is also what fuels the popular media trope of “AI takeovers.” But at this point, it’s all speculation.

“Artificial superintelligence will become by far the most capable forms of intelligence on earth,” said Dave Rogenmoser, CEO of AI writing company Jasper. “It will have the intelligence of human beings and will be exceedingly better at everything that we do.”

More on AI4 Types of Machine Learning to Know

Functionality-Based Types of Artificial Intelligence

Functionality is focused on how an AI applies its learning capabilities to process data, respond to stimuli and interact with its environment. As such, AI can be sorted by four functionality types.

4. Reactive Machine AI

Reactive machines are just that — reactionary. They can respond to immediate requests and tasks, but they aren’t capable of storing memory, learning from past experiences or improving their functionality through experiences. Additionally, reactive machines can only respond to a limited combination of inputs. Reactive machines are the most fundamental type of AI.

In practice, reactive machines are useful in basic autonomous functions, such as filtering spam from your email inbox or recommending items based on your shopping history. Beyond that, reactive AI can’t build upon previous knowledge or perform more complex tasks.

Reactive Machine AI Examples

• IBM Deep Blue: IBM’s reactive AI machine Deep Blue was able to read real-time cues to beat Russian chess grandmaster Garry Kasparov in a 1997 chess match. 
• Netflix Recommendation Engine: Media platforms like Netflix often utilize AI-powered recommendation engines, which process data from a user’s watch history to determine and suggest what they would be most likely to watch next.

5. Limited Memory AI

Limited memory AI can store past data and use that data to make predictions. This means it actively builds its own limited, short-term knowledge base and performs tasks based on that knowledge.

The core of limited memory AI is deep learning, which imitates the function of neurons in the human brain. This allows a machine to absorb data from experiences and “learn” from them, helping it improve the accuracy of its actions over time. 

Today, the limited memory model represents the majority of AI applications. It can be applied in a broad range of scenarios, from smaller-scale applications, such as chatbots, to self-driving cars and other advanced use cases.

Limited Memory AI Examples

• Chatbots and Virtual Assistants: Chatbots and virtual assistants are forms of limited memory AI that use deep learning to mimic human conversation. As users interact more with these systems, they learn from this data and remember details about the user, allowing them to provide relevant and personalized responses.
• Self-Driving Cars: Self-driving cars continually observe and process environmental data around them as they travel on the road. This helps them predict when they need to turn, stop or avoid an obstacle. 

6. Theory of Mind AI

Theory of mind in AI refers to the ability to recognize and interpret the emotions of others. The term is borrowed from psychology, describing humans’ ability to read the emotions of others and predict future actions based on that information. Although it has not been achieved yet, theory of mind would be a substantial milestone in AI’s development. 

An emotionally intelligent AI could bring a lot of positive changes to the tech world, but it also poses some risks. Since emotional cues are so nuanced, it would take a long time for AI machines to perfect reading them, and could potentially make big errors while in the learning stage. And some worry that an AI capable of responding to both emotional and situational signals could lead to the automation of more jobs.

7. Self-Aware AI

Self-aware AI refers to the hypothetical stage of artificial intelligence where machines possess self-awareness. Often referred to as the AI point of singularity, self-aware AI represents a stage beyond theory of mind and is one of the ultimate goals in AI development. It’s thought that once self-aware AI is reached, AI machines will be beyond our control, because they’ll not only be able to sense the feelings of others, but will have a sense of self as well. 

Additional Types of Artificial Intelligence

AI technologies are also developed for specific purposes and use cases. While these types of AI are more application-based, they’re still helpful to know.     

AI in Robotics

AI is improving the field of robotics in various ways. AI powers machine vision, giving robots the ability to identify objects and navigate different environments on their own. Robots can also perform repetitive tasks with the help of AI, such as picking produce on farms, transporting packages in warehouses and identifying medical images that reveal disease risk factors in healthcare settings. 

The next generation of robots will depend on AI to function, too. Collaborative robots possess a system of sensors and advanced functions that enable them to remain aware of their surroundings and engage with human workers safely. AI is also leading to more general-purpose robots, which can understand verbal commands and learn new tasks independently. 

AI in Robotics Examples

• Figure AI: Figure AI is building bipedal humanoid robots that are designed to work alongside humans and reason just like humans do. So far, Figure AI has built Figure 01 and Figure 02, with the goal of eventually addressing labor shortages.    
• Tesla: Tesla has introduced its own humanoid robot known as Optimus. Like Figure AI’s robot, Optimus is intended to be a general-purpose robot and has demonstrated the ability to traverse uneven terrain. 

Computer Vision

Computer vision gives AI technologies the ability to process visual information and convert it into usable data. This makes it possible for AI software, robots and other machines to detect objects, track moving objects and map out a physical environment, among other applications.

A subset of computer vision known as image recognition offers even more possibilities. Referring to the process of identifying and classifying different elements within an image, image recognition supports many use cases like image-based medical diagnoses, security systems and inventory management. 

Computer Vision Example

• An everyday example of computer vision is facial recognition. This technology can analyze and store data on a person’s facial features, essentially identifying a person based on their face. As a result, facial recognition has become a convenient security feature in smartphones, allowing a user to unlock their phone with their face. 

Expert Systems

Expert systems are trained on data sets to solve complex problems using rule-based decision-making processes. AI expert systems are trained either through forward chaining or backward chaining. In forward chaining, a system starts with facts and learns how to make inferences to gain more information until a goal is achieved. In backward chaining, the system starts with the goal and works backward to determine the facts used to reach the goal. 

Expert Systems Example

• MYCIN was an early expert system intended to help diagnose bacterial infections and advise physicians on choosing the appropriate treatment. The computer program relied on backward chaining, revisiting its rules and facts to understand its own reasoning and explain its decisions to users.

Frequently Asked Questions