An artificial intelligence system or AI system is either a physical machine or a computer program that can perform tasks that require human intelligence. Developing a particular system involves a combination of an end-to-end design called architecture, a set of instructions or rules called algorithms, and a representation of a real-world phenomenon called a model.
There are different examples of AI systems. Most have been deployed while some remain hypothetical, Furthermore, depending on their learning methods, the intended application, or stated goals, there are also different ways to categorize a particular AI system can be classified. This article discusses the major types of artificial intelligence systems.
Kaplan and Haenlein Classification: Three Types of Artificial Intelligence Systems Based on Intellectual Capabilities
Researchers Andreas Kaplan and Michael Haenlein borrowed concepts from the management literature to introduce three broad categories of artificial intelligence systems. These are analytical artificial intelligence, human-inspired artificial intelligence, and humanized artificial intelligence. The following are further details of each:
1. Analytical Artificial Intelligence
An analytical AI system has characteristics similar to and consistent with cognitive intelligence. This means that it has the capabilities to acquire knowledge and demonstrate an understanding of this knowledge through learning or training and inference.
This particular system is trained using a representation of the world and can produce inferences using its knowledge of the world. Some can learn from experience or through feedback. A machine learning model trained on a large dataset is an example.
Specific applications of an analytical AI system are image recognition software, autonomous vehicles, and intelligent chatbots and virtual assistants. It is worth mentioning that analytical AI systems are the most common artificial intelligence systems.
2. Human-Inspired Artificial Intelligence
A human-inspired AI system combines the elements of cognitive intelligence and emotional intelligence. This equips it with both analytical capabilities based on its knowledge of the world and capabilities to process and understand human emotions.
The possible applications of this system center on assessing human emotions to aid in making decisions. A specific example is a software that can look for cues from facial expressions using facial recognition or patterns in intonation using voice recognition.
Walmart has installed cameras in some of its stores that can pinpoint frustrated customers. This allows store personnel to provide suitable customer service. The Swedish company Smart Eye has developed a software for determining dangerous driving behavior.
3. Humanized Artificial Intelligence
Another type of artificial intelligence system is called humanized AI system. It has cognitive intelligence and emotional intelligence while factoring social intelligence into its overall capabilities. This system is self-aware and environment-aware.
Technological limitations still render humanized artificial intelligence systems hypothetical. However, once realized, their purported applications center on the development and deployment of intelligent and self-aware AI agents and autonomous AI assistants.
Some of the imagined examples of a humanized AI system include virtual customer support representatives that can deal with customer concerns, computer-powered autonomous teachers, and virtual companions that can provide entertainment and support.
Four Modern Classifications: Four Types of Artificial Intelligence Systems Categorized Based on Scope of Capabilities
Another criterion for categorizing different artificial intelligence systems is the scope of their capabilities. Hence, unlike the classification based on intellectual capabilities such as cognitive, emotional, and social intelligence, these categories represent the extent of the tasks these systems can perform. Below are more specific details:
1. Reactive Machines
The simplest type of artificial intelligence system falls under the reactive machines classification. These systems do not have learning and memory capabilities nor they can use past experiences to guide their actions or decisions. Their response to situations is limited to a predefined set of rules. Examples of reactive machine AI systems include video games such as chess and classic strategy-based board games and robotic vacuum cleaners.
2. Limited Memory Machines
Limited memory machines are more advanced versions of reactive machines is called. These AI systems can make more sophisticated decisions using their stored memories or past experiences. However, despite these capabilities, they do not have long-term memory and a sense of self. Examples include facial recognition systems and autonomous vehicles. These examples can perform tasks that require short-term observations.
3. Theory of Mind Machines
There are hypothetical artificial intelligence systems that can understand and interact with other agents such as humans and even similar AI systems. Take note that these agents have their own beliefs, values, desires, preferences, and intentions. A theory of mind machine AI system can interface with these agents. It is an intelligence and sociable system that can model emotions and social behaviors or communication using natural language.
4. Self-Aware Machines
Another type of AI system that can demonstrate a conventional level of self-awareness is called a self-aware artificial intelligence. It is both autonomous and aware of its own existence and place in the world. It can reflect on its own actions and goals or objectives, or even develop its own purpose and make decisions based on its personal values and preferences. This AI system remains hypothetical because of technological limitations.
Levelling-Based Classification: Three Types of Artificial Intelligence Systems Based on Level of Intelligence
There are also types of artificial intelligence systems categorized according to their intelligence level. This classification model helps in communicating to the public the progress in the overall field of AI and keeping track of related developments or understanding in which level a new system falls under. Take note of the following:
1. Artificial Narrow Intelligence or Weak Artificial Intelligence
An artificial narrow intelligence or ANI and weak AI is an AI system that can perform specific tasks well but still lacks the capabilities to generalize its knowledge to other non-specific tasks. This is the current level of progress in artificial intelligence. The most common examples are the use of deep learning to train large language models for generative AI applications such as chatbots or computer vision models for advanced image recognition.
2. Artificial General Intelligence or Strong Artificial Intelligence
The next level of progress in artificial intelligence is the arrival of artificial general intelligence or AGI and strong AI. A particular AGI system has a level of understanding and reasoning that is comparable to humans. It can also use its knowledge to a wide range of tasks and its applications would not be limited to a single or limited domain. Advances in large and complex AI models are seen to draw the field closer to achieving this level.
3. Artificial Superintelligence or Super Artificial Intelligence
An artificial superintelligence or ASI represents hypothetical artificial intelligence systems that are more intelligent than humans. It would be able to surpass human intelligence and outperform human intellectual capabilities in all respects to enable it to solve problems that are beyond the reach of human cognition. An ASI can spur a technological singularity characterized by an uncontrollable and irreversible intelligence explosion.
Practical Classifications: Main Types of Artificial Intelligence Systems Based on Practical AI Subfield Applications
There are different goals within the general field of artificial intelligence. Some of these goals correspond to the subfields of artificial intelligence. The practical applications of these subfields have also been used to categorize different artificial intelligence systems. Examples include machine learning and deep learning systems, natural language processing systems, artificial neural networks, and computer vision systems.
FURTHER READINGS AND REFERENCES
- Kaplan, A. and Haenlein, M. 2019. “Siri, Siri, In My Hand: Who is the Fairest in the Land? On the Interpretations, Illustrations, and Implications of Artificial Intelligence.” Business Horizons. 62(1): 15-25. DOI: 1016/j.bushor.2018.08.004
- Liu, B. 2021. “Weak AI” is Likely to Never Become “Strong AI”, So What is Its Greatest Value for Us?” arXiv. DOI: 48550/ARXIV.2103.15294
- Russel, S. and Norvig, P. 2009. Artificial Intelligence: A Modern Approach. 3rd Ed. Pearson. ISBN: 978-0136042594
- Tauli, T. 2019. Artificial Intelligence Basics: A Non-Technical Introduction. Apress. DOI: 1007/978-1-4842-5028-0