Early iterations of the AI applications we use most today were built on traditional machine learning models. These models rely on learning algorithms that are developed and maintained by data scientists. In other words, traditional machine learning models require human intervention to process new information and perform new tasks that were not covered by initial training. For example, Apple made Siri a feature of its iOS in 2011. This early version of Siri was trained to understand a very specific set of statements and requests. Human intervention was required to expand Siri's knowledge base and capabilities.
However, since the breakthrough development of artificial neural networks in 2012, AI capabilities have steadily evolved. This allows machines to perform reinforcement learning and simulate the way the human brain processes information. Unlike basic machine learning models, deep learning models allow AI applications to learn how to perform new tasks that require human intelligence, perform new actions, and make decisions without human intervention. You can learn how. As a result, deep learning is enabling task automation, content generation, predictive maintenance, and other capabilities across industries.
The field of AI is still rapidly changing due to deep learning and other advances. Our collective understanding of realized and theoretical AI continues to change, meaning that AI categories and AI terminology may differ (or overlap) from one source to another. However, the types of AI can mostly be understood by considering two overarching categories: AI capabilities and AI capabilities.
3 types of AI based on ability
1. Artificial narrow-area AI
Narrow artificial intelligence, also known as weak AI, what we call narrow AI, is the only type of AI that currently exists. Other forms of AI are theoretical. It can be trained to perform single or narrow tasks, and is often much faster and better than the human brain. However, it cannot be executed outside of the defined task. Instead, they target a single subset of cognitive abilities and progress within that range. Siri, Amazon's Alexa, and IBM Watson are examples of Narrow AI. Even OpenAI's ChatGPT is considered a type of Narrow AI because it is limited to a single task: text-based chat.
2. General AI
Artificial general intelligence (AGI), also known as strong AI, is just a theoretical concept today. AGI can use previous learning and skills to perform new tasks in different contexts without requiring humans to train the underlying model. This ability allows AGI to learn and perform any intellectual task humanly possible.
3.Super AI
Super AI is commonly referred to as artificial superintelligence, and like AGI, it is strictly theoretical. If realized, super AI would think, reason, learn, and make decisions, and would have cognitive abilities that surpass those of humans. Applications with super AI capabilities will evolve beyond understanding human emotions and experiences to feel emotions, have needs, and have their own beliefs and desires.
4 types of AI functions
There are two functional AI categories under Narrow AI, which is one of three types based on functionality.
1. Reactive Machine AI
Reactive machines are memoryless AI systems designed to perform very specific tasks. We only use currently available data because we cannot remember previous results or decisions. Reactive AI originates from statistical mathematics and can analyze vast amounts of data to produce seemingly intelligent output.
Example of reactive machine AI
- IBM deep blue: IBM's competitive chess supercomputer AI defeated chess grandmaster Garry Kasparov in the late 1990s by analyzing the pieces on the board and predicting the likely outcome of each move.
- Netflix Recommendation Engine: Netflix viewing recommendations utilize models that process data sets collected from viewing history to deliver content that customers are most likely to enjoy.
2. Limited memory AI
Unlike Reactive Machine AI, this form of AI remembers past events and outcomes and can monitor specific objects or situations over time. Limited Memory AI can use past and current data to determine the course of action most likely to help achieve a desired outcome. However, while Limited Memory AI can use past data for a period of time, it cannot keep that data in a library of past experiences for long-term use. Limited Memory AI's performance improves as it is trained with more data over time.
Limited memory AI example
- Generation AI: Generative AI tools like ChatGPT, Bard, and DeepAI rely on limited memory AI capabilities to predict the next word, phrase, or visual element in the content they generate.
- Virtual assistants and chatbots: Siri, Alexa, Google Assistant, Cortana, and IBM Watson Assistant combine natural language processing (NLP) and memory-constrained AI to understand questions and requests, take appropriate actions, and craft responses.
- self-driving car: Self-driving cars use Limited Memory AI to understand the world around them in real time and make informed decisions like speeding up, braking, or turning.
3. AI theory of mind
Theory of Mind AI is a functional class of AI that falls under general purpose AI. Although it is a form of AI that is not available today, an AI with theory of mind capabilities would understand the thoughts and emotions of other beings. This understanding can influence how AI interacts with the people around it. In theory, this would allow AI to simulate human-like relationships. Theory of Mind AI can infer human motivations and reasoning, thus personalizing interactions with individuals based on their unique emotional needs and intentions. Theory of Mind AI can also understand and contextualize artwork and essays, something that today's generative AI tools cannot do.
Emotional AI is an AI theory of the mind currently under development. AI researchers hope that AI will be able to analyze audio, images, and other types of data to recognize, simulate, and monitor humans at an emotional level and respond appropriately. Currently, emotional AI is unable to understand and respond to human emotions.
4. Self-aware AI
Self-Aware AI is a type of functional AI class for applications with super AI capabilities. Like AI theories of mind, self-aware AI is strictly theoretical. If realized, it would be possible to understand human emotions and thoughts, as well as our own internal states and characteristics. You may also have your own feelings, needs, and beliefs.
Emotional AI is a theory of mind AI currently under development. Researchers hope it will have the ability to analyze audio, images, and other types of data to recognize, simulate, and monitor humans at an emotional level and respond appropriately. Currently, emotional AI is unable to understand and respond to human emotions.
Additional features and practical applications of AI technology
computer vision
Narrow AI applications with computer vision can be trained to interpret and analyze the visual world. This enables intelligent machines to identify and classify objects in images and video footage.
Applications of computer vision include:
- Image recognition and classification
- object detection
- object tracking
- face recognition
- Content-based image search
Computer vision is important for use cases where AI machines interact with or traverse the physical world around them. Examples include self-driving vehicles and machinery moving through warehouses and other environments.
robotics
Robots in industrial environments can use Narrow AI to perform routine, repetitive tasks such as material handling, assembly, and quality inspection. In the medical field, robots equipped with Narrow AI can help surgeons monitor vitals and detect potential problems during procedures. Agricultural machinery can autonomously prune, move, thin, sow, and spray. Smart home devices such as the iRobot Roomba can also use computer vision to navigate the interior of your home and understand its progress using data stored in its memory.
expert system
Expert systems with Narrow AI capabilities can be trained on corpora to emulate human decision-making processes and apply their expertise to solve complex problems. These systems can evaluate large amounts of data to uncover trends and patterns and make decisions. It also helps businesses predict future events and understand why past events occurred.
IBM's AI
IBM has been a pioneer in AI since the beginning, delivering breakthrough after breakthrough in the field. IBM recently released a major upgrade to its cloud-based generative AI platform known as watsonx. IBM watsonx.ai brings together new generative AI capabilities that leverage foundational models and traditional machine learning into a powerful studio that spans the entire AI lifecycle. watsonx.ai allows data scientists to build, train, and deploy machine learning models in a single collaborative studio environment.
Explore watsonx.ai now
Was this article helpful?
yesno
