Exploring the types of AI: Reactive machines to superintelligence

AI: Friend or foe?

People have mixed feelings about artificial intelligence (AI). For some, the term conjures the chilling image of a leather-clad Arnold Schwarzenegger hunting down humans. For others, they imagine the pioneering NASA rovers taking photographs and collecting data on Mars. Despite their drastic differences, both are relevant examples.

What is artificial intelligence?

One of the things that makes this technology so potent is that AI is not limited to one form or another. AI enables computers and machines to simulate human learning, comprehension, problem-solving, decision-making, creativity, and autonomy. Applications and devices equipped with AI can see and identify objects, even understand and respond to human language.¹ Other popular examples of AI include self-driving cars, virtual assistants like Amazon’s Alexa, and generative tools like ChatGPT. The tremendous power and potential of AI could end up changing all elements of life and how we do business—in a dramatically similar fashion to the advent of the internet.

The evolution of AI

While its origins can be traced back to the 1950’s, AI technology remained in its infancy for several decades. Toward the end of the 1980’s, thanks to renewed interest and an increase in research and development funding, AI took massive leaps forward.² With automation and robotics becoming an efficient solution for many businesses at the beginning of the 21^st century, the groundwork was laid for AI to revolutionize entire industries. A major breakthrough occurred in 2012 with the development of artificial neural networks, which allow AI models to engage in reinforcement learning and simulate how the human brain processes information.³

Data scientists and computer programmers endeavored to create smarter, faster iterations, and fascination with AI grew. Combined with robotics, AI can be used to take vital signs from patients in nursing homes. When linked to a thermostat, it can monitor a house’s temperature and adapt to fluctuating environmental conditions. But not all forms of AI inhabit a physical model; there are different types of AI used for different purposes.

Machine learning and other types of AI

Before detailing the various types of AI, it’s important to first understand machine learning (ML). Relying on algorithms and datasets developed by scientists, ML is a prerequisite of artificial intelligence that’s enabled through models that can “learn” from data patterns without human direction.⁴

So, the more exposure they have to a certain dataset, the more adept they are at this pattern recognition. As machines become increasingly savvy at recognizing those patterns, they develop capabilities that inform future interactions and responses. While this technology has been widely adopted for its efficiency benefits, it still requires human input to provide the material the machines learn from.

Based on how they learn and how extensively they can apply their knowledge, all AI can be grouped into categories, based on capabilities or functionalities.⁵

Types of AI based on capabilities

Narrow AI

Also known as artificial narrow intelligence, this describes AI tools designed to execute specific tasks or commands. They’re designed for one cognitive capability and can’t learn independently beyond what they’re designed for. All the AI systems used today fall under this category, with common examples being large language models (LLMs), voice assistants, and customer service chatbots.

General AI

This type of AI is theoretical and refers to a future form of AI that can learn, think, and perform a large variety of tasks at a level equivalent to humans. This AI would leverage previous skills to complete new tasks—without the need for human input to train them first.

Super AI

The stuff of science fiction novels and nightmares, artificial superintelligence, or super AI, would theoretically be able to learn at such an accelerated rate that its knowledge and abilities would soon outstrip that of humans. These hypothetical, advanced capabilities being used against our species is part of what fuels the “robot takeover” paranoia.

Types of AI based on functionalities

Reactive machines

One of the most fundamental forms of AI, reactive machines are capable of immediate response (to a limited number of commands), but they cannot store memories or learn from previous experiences. Their self-explanatory name illustrates their main function—to react. Notable examples of reactive AI include IBM’s chess-playing supercomputer, Deep Blue. This AI beat a chess grandmaster in the late 1990’s through chess board analysis and predicting the probability and outcomes of each move.⁶

Limited memory

Differing from reactive machine AI, limited memory AI can recall previous experiences and observe certain objects or data over a period of time. They can also use data from the present moment to determine a problem-solving approach. This recall is limited to the short term; the AI can’t keep a library of information to draw from. Performance of the AI can improve with more data exposure and longer training sequences.

Generative AI and LLMs are prime examples of limited memory AI. Through neural networks trained on large amounts of text data, LLMs use predictive behavior to generate text in response to a prompt. Other forms of generative AI can create human-like content, from pictures and videos to poetry and even computer code.⁷

Theory of mind

Another purely theoretical form of AI, theory of mind fits within the general AI category, able to think, learn, and understand the thoughts and emotions of others on par with humans. This creates the possibility that AI can engage in human-like relationships. Because theory of mind AI could infer human motives and reasoning, it would tailor its interactions with people based on its interpretation of their needs.

An often-referenced example of theory of mind AI is Jarvis from the Iron Man comics and movies. Jarvis was a hyper-capable version of Amazon’s Alexa, able to take input from a seemingly unlimited number of sources and execute very complicated analysis, but it was also endowed with what appeared to be a personality. The AI had an enormous amount of contextual awareness about every human who interacted with it, allowing it to anticipate and interpret user needs without being prompted. Jarvis likely represents the height of expectation-setting for AI capability: able to do nearly anything at incredible speeds, stopping just short of self-awareness.

Self-aware

Similar to super AI, self-aware AI is exactly how it sounds; it's aware of its own traits, conditions, “identity”, etc. This purely theoretical form of AI would possess consciousness and have the ability to perceive its own existence.⁸ If it was ever achieved, there are significant implications (both positive and negative) for how this AI could affect humans. If the AI realized it had superseded human intellect and ability, what would happen next? If you ask the more cynical sci-fi fans, this marks the beginning of the “robot takeover.”

The future of AI

As adoption of AI continues, and use cases become more complex, we’ll continue to find opportunities for drastic leaps in innovation, as well as massive risks and pitfalls. What will become paramount to our relationship with AI is the responsible and ethical use of the technology.

Imagining a future without AI is nearly impossible; it’s also improbable, since human beings have a fairly consistent track record. As soon as we develop the next best thing, whether it’s phones or computers, there is no going back to the early models. An illustration of this evolution would be the telephone—not many people use landlines anymore, now that smartphones have become the norm. Even with smartphones, every year there are new and improved models being released, and very few people continue to use the older versions. It’s likely that the future of AI, and how we use it, will be similar.

Responsible AI deployment is not just about technological prowess but also about ethical considerations and robust governance frameworks. At CAI, we help organizations identify and assess opportunities for AI deployment and expansion, working as a trusted partner mitigate risks and maximize the benefits of AI.

To learn more about how CAI empowers businesses with responsible AI, fill out the form below.