Artificial Intelligence (AI) is a complex and diverse field. In computer science, it refers to numerous technologies and frameworks that encompass an ever-evolving area of specialty. Focusing on creating machines and systems that need no human intervention to perform tasks and make decisions, understanding AI can be overwhelming. Machine learning, speech recognition assistants such as Alexa and Siri, and even the algorithm that recommends movies for you on Netflix are some of the examples we encounter daily. As advances in AI continue to profoundly impact both business and society, understanding this field of computer science can help you leverage its untapped potential.
Artificial Intelligence Classification
Artificial Intelligence is an all-encompassing term, and understanding this field of study can be both confusing and overwhelming. One way of determining and carving AI into a workable framework that your mind can process is to classify it by type.
Reactive AI defines systems with no memory that only respond to identified internal or external events. This classification describes the oldest form of AI. Solutions that fall into this category have limited autonomous capability. These systems or machines only execute an action when a process breaches a particular threshold, or an event triggers a specific response. One could argue that reactive AI existed before computers. For example, a steam pressure release valve that automatically opens to ensure a boiler’s safety is a reactive, unattended system that responds to an event.
There are many examples of reactive AI systems in modern computing. Monitoring systems that trigger alerts and send emails when devices breach certain thresholds are an excellent example. Other reactive AI solutions include chess-playing computers and other gaming platforms. They do not have any memory, and their developers program them to react when they encounter a particular event.
Limited Memory AI
Computer scientists classify the next evolution in Artificial Intelligence as limited memory. Systems that fall into this category utilize memory and experience to learn and improve their responses. Machine Learning (ML) technology is an excellent example of this classification. Leveraging various algorithms, ML uses the knowledge gained from past experiences and access to existing knowledge bases to make independent decisions.
There are many examples of limited memory AI that we all interact with either directly or indirectly daily. The example most often cited is self-driving vehicles. Their decision-making processes use both reactive AI elements as well as data collected recently. This area of AI has grown exponentially in recent years due to the big data and deep learning algorithms made possible by the rise of cloud computing.
Theory of Mind
The Theory of Mind classification is purely academic at this time. As computer scientists have not created or implemented systems with this capability yet, the hypotheses found in this categorization are currently conceptual. The essence of this area is Artificial Intelligence that understands and remembers emotions in other intelligent entities. Based on this deeper interactive understanding, it can then adjust its behavior accordingly. This form of AI’s core premise is that it will genuinely comprehend the emotional state and then interact in a socially acceptable manner. The key to unlocking this potential is creating AI that perceives human beings as individuals. Many computer scientists believe this requirement will spawn other AI theory branches if this conceptual area becomes a reality.
The fourth and final AI classification is Self-Aware. Like Theory of Mind, this categorization is purely theoretical at the moment. This next step in the evolutionary machine process will create systems and devices that understand and react to inner emotions. This level of self-awareness will develop machines with human-like intelligence and understanding. The theory is that these sentient systems will possess the same innate desires, needs, and emotions as human beings. Unlike the Theory of Mind, where machines will treat each individual appropriately, Self-Aware AI will create systems that equal or surpass their human creators.
Classifying Artificial Intelligence by Role or Application
Another way to view and classify the various Artificial Intelligence types is to categorize them by role or application. This broad categorization refers to all AI classification types from Reactive and Limited Memory solutions that exist today to the Theory of Mind and Self-Aware AI solutions of the future.
Assisted Artificial Intelligence
Assisted Intelligence is a basic AI level where automation helps with repetitive mundane tasks and monotonous procedures. Robots on an assembly line and systems that help fill logistical orders are good examples of this technology type.
Augmented Artificial Intelligence
Augmented Intelligence refers to technologies that support and supplement rather than replace human understanding. Good examples of Augmented Intelligence are financial AI algorithms that help predict stock and foreign exchange price movements based on historical data.
Autonomous Artificial Intelligence
Autonomous Intelligence relates to systems that can act without human intervention. Although this definition may sound like science fiction, there are numerous examples of this type of AI. Technology monitoring solutions that automatically restart services are autonomous systems. Another example could be the web scraping tools that search for particular words that trigger subsequent workflows.
The Artificial Intelligence Spectrum
A third and final method to understand the Artificial Intelligence landscape is to see it as an endless spectrum from basic or narrow AI to superintelligence.
Artificial Narrow Intelligence (ANI)
Artificial Narrow Intelligence represents all existing AI. This classification refers to AI solutions that can perform a particular task a human would ordinarily perform. These systems or devices cannot do anything not explicitly stated in their programming. Even today’s most complex machine learning algorithms fall under this category.
Artificial General Intelligence (AGI)
Artificial General Intelligence is an AI agent with the capability to learn, understand, and function like a human being. When AI reaches this maturity level, it will consist of systems that can form connections across knowledge domains. With this information, AI platforms will then have the ability to make independent decisions with no direct or indirect human input. The primary difference between Narrow and General AI is the level of training required and their range of capability. Unconstrained, AGI can perform various tasks across domains simultaneously. In contrast, ANI is limited to a particular programmed function and environment.
Artificial Superintelligence (ASI)
Artificial Superintelligence refers to AI expected to become the pinnacle of earth-based intelligence. Over and above having the AGI abilities of a human being, ASI entities will possess more significant memory and instantaneous access to data far surpassing human limits. Having machines with this capability to answer questions while performing an unlimited number of activities will exceed even the most remarkable human capabilities. However, if these systems are sentient, computer scientists fear the consequences should they turn against their creators. Ultimately, this type of AI will depend on the boundaries human beings set to achieve the productivity they require while ensuring their safety and security.
Classifications and Continuums
Artificial Intelligence is an endless field of study, and trying to grasp the various terms and traits that make up an AI system can be complicated. Using a combination of classification types plotted on an endless spectrum, one can then define the appropriate boundaries needed for the safe creation and consumption of this technology.