What are the types of artificial intelligence (AI) being used?

• Intro

• Different types of AI by capabilities

• Artificial General Intelligence (AGI)

• Artificial Narrow Intelligence (ANI)

• Artificial Superhuman Intelligence (ASI)

• Different types of AI by functionalities

• Limited Memory AI

• Reactive Machines AI

• Self Aware AI

• Theory of Mind AI

• Conclusion

Artificial Intelligence (AI) is integrated into nearly every aspect of our lives. AI is generally defined as the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. Another way to put it is the simulation of human intelligence processes by machines and computer systems from large amounts of data. Innovation through AI can be found in almost every business sector (e.g.

,

,

,

,

,

,

,

,

,

, etc) but there are different types of AI divided by its few capabilities and functionalities. We’ll go over each type of AI and provide examples of how they are used today.

Artificial General Intelligence, also known as strong AI, full AI, or general intelligence action works with the ability to learn, perceive, understand, and function just like a human being.

Such systems will have the capability to independently build multiple competencies and form connections and generalizations across domains, significantly reducing the time required for training.

Couple of examples of AGI include:

• Pillo Health

  robot is a countertop automated medication management and voice first smart home healthcare device. It can remind users and dispense medication on a pre-set schedule. Pillo can also deliver audio or video-based health content and facilitate video calls between patients and their caregivers, while allowing them to monitor adherence data from the device.

• Fujitsu

  built the K computer, the world’s first supercomputer that broke the 10

  petaflops

  barrier. It’s 1 of the closests attempts at achieving strong AI, but taking nearly 40 minutes to simulate a single second of neural activity, has raised questions of if strong AI will be achieved in the near future.

Artificial Narrow Intelligence, also known as weak AI, or general purpose AI refers to AI systems that can only perform a specific task autonomously using human-like capabilities. They do what they’re specifically programmed to do which is a limited (narrow) range of competencies. Even the most complex AI that uses machine learning (ML) and deep learning to teach itself can fall under the category of ANI.

Few examples of AGI include:

• Apple Siri

  operates with a limited pre-defined range of functions and often has problems with tasks outside its general capabilities.

• IBM Watson

  supercomputer is another example of a Narrow AI. With its cognitive computing, machine learning (ML), and natural language processing (NLP) to process information and answer queries. IBM Watson once out-performed human contestant Ken Jennings to become the champion on the popular game show, Jeopardy!

• Other examples of Narrow AI include

  Google Translate

  ,

  Google’s page-ranking algorithm

  , image recognition softwares, recommendation systems, and spam filtering systems.

Now let’s get to the good stuff. Artificial Superhuman Intelligence also known as superhuman AI, or superintelligent AI are systems that surpass human intelligence and have the capability to perform tasks better than a human. ASI will be exceedingly better at everything they do because of its greater memory, faster data-processing and analysis, and decision-making capabilities. The further development of AGI and ASI has led to controversial concepts called singularity which is a hypothesis that technological growth may one day become uncontrollable or irreversible.

For now, this is merely a concept and hard to picture when more advanced types of AI will come to fruition, as most AI still fall under the ANI category and still in its rudimentary stage. For those who have concerns and potentially a negative outlook on the future of AI, rest assured that it is still too early to be losing sleep over singularity and there’s time to ensure we take safety precautions in AI development.

Limited Memory AI trains and learns from historical data to make decisions. The past data is used for a specific period of time causing the memory of the system to be short-lived and often cannot be added to a library of their experiences. Nearly all present-day AI systems fall under this AI.

Few examples of Limited Memory AI include:

• Tesla’s Full Self-Driving (FSD) is an example of limited memory AI in today’s world. Movements of vehicles are constantly detected and static data such as lane markers, traffic lights, and curves on the road are added to the AI machine to avoid nearby vehicles and prevent accidents.

• Other examples include

  chatbots

  and

  virtual assistants

  and pretty much all modern-day AI applications.

Reactive Machines AI are the oldest forms of AI systems with extremely limited capabilities. This is a primary form of AI that does not store memories or use past experiences to determine future actions, only using present data. Therefore, it is reactive and can only be used to automatically respond to a limited set or combination of inputs for specific tasks.

Few examples of Reactive Machines AI include:

• IBM’s chess-playing computer,

  Deep Blue

  , is an example of reactive machine AI. In the late 1990s, Deep Blue defeated international grand-master, Garry Kasparov, in chess. Only identifying current pieces on a chessboard, Deep Blue makes next move predictions, ignoring all prior experiences.

• In a similar way, Google’s

  AlphaGo

  , defeated the top human Go experts but more enlightened than Deep Blue by using a neural network to assess the game development.

Self Aware AI, a supplement of the theory of mind AI, is the final stage of AI development and only exists hypothetically. They possess human-like consciousness and reactions and will be able to understand and evoke emotions in those it interacts with, as well as have emotions, needs, and beliefs of its own. Development of self-awareness can lead to a boost in progress to civilization but others are worried of the implications that could potentially lead to harmful outcomes pertaining to self-preservation. It is believed to be decades, if not centuries away from materializing, but remains the ultimate objective of AI research.

Theory of Mind AI is considered very advanced technology and like Self Aware AI, only exists as a concept and a work in progress. Theory of mind level AI will be able to better understand the entities it is interacting with by discerning needs, emotions, beliefs, and thought processes. In order to achieve theory of mind AI, further development in other branches of AI will be required and in order to essentially become “understanding” humans — a thorough understanding that people and things within an environment can alter feelings and behaviors.

Real-world examples:

• Kismet

  is a robot head created in the late 90s by a Massachusetts Institute of Technology (MIT) researcher. Kismet can mimic and recognize human emotions. Though considered key advancements in theory of mind AI, Kismet can’t follow gazes or convey attention to humans.

• Sophia

  from Hanson Robotics is another example of theory of mind AI. Cameras in Sophia’s eyes, combined with computer algorithms, allow her to see, sustain eye contact, recognize individuals, and follow faces.

AI depicted in society often doesn’t match the modern day reality. It’s important to acknowledge the different forms and levels of AI technology to understand the advantages and benefits of AI, while reducing the skepticism and fear that can sometimes surround it. From your everyday AI used today (Limited Memory, Reactive Machines) to building towards the conceptual AI (Theory of Mind, Self Aware) of tomorrow, AI is at the heart of transforming society with the end-goal of making human lives better and easier.

Any requests for future Q&As? Contact us at

.