Capability Levels
AI has been classified into mutiple levels based on capability and functionality. We are listing the commonly used classifications below.
While we are listing the general understanding of levels below, there are no hard and fast rules when classifying any solution into a level. The boundaries are often blurred and can vary by context. We leave it to you to follow the trending conversations and identify what your solutions may be categorized as.
Capability by Scope
1. Narrow or Weak AI
Solutions at this level of AI can be trained to perform a single or a few tasks within a narrow scope faster and more efficiently, accurately, and consistently than a human mind can. They can be used to automate routine tasks and help in decision-making.
Solutions at this level are the most commonly used and some have existed for a while in various forms known as Decision Support Systems and Robotic Process Automation.
2. General or Strong AI
In addition to narrow AI functions, solutions at this level are designed to think, reason, make decision, and act autonomously, without requiring any human intervention. They are also designing to learn from past data and action, and can multiple tasks in a wide scope and context.
This is an advanced level of AI that is in an emerging state and while there are some solutions that work at this level, they are still evolving.
3. Super AI
Super AI solutions would be capable of outperforming humans by working at the level of general AI and also understanding and interpreting human emotions and experiences.
This is yet an experimental level of AI and stable solutions for general use are yet to be developed.
Capability by Design and Function
4. Reactive Machines
Solutions of this type are the most basic forms of AI. They are essentially programmed to be able to respond to a set of fixed inputs and provide logical output. they may be able to combine vairous inputs and take decisions based on the various permutations and combinations. They do not have the ability to remember past data, decisions, or actions, and cannot change or improve upon the output by learning.
Even being the most basic form of AI they would still be able to handle more input data, and handle more permutations and combinations, and provide faster, more accurate, and consistent decisions and outputs than humans.
5. Limited Memory AI
Limited memory solutions have the capabilities of Reactive Machines and are also capable of learning from past data, decisions, or actions, and can change or improve upon the output by learning. Such solutions progressively improve their outputs and decisions over time as they get access to increasing amounts of data, present and past (adding the Machine Learning dimension to AI).
6. Theory of Mind AI
Theory of Mind AI solutions would have the capabilities of Limited Memory AI solutions and in addition be able to read, sense, and understand the thoughts and emotions of humans they are interacting with. This capability would allow solutions to deliver different outputs and decisions based on the current environment, context, or even the user’s state of mind (based on speech patterns for example). In theory, this would allow solutions to simulate dynamic human-like interactions.
7. Self-Aware AI
Self-Aware AI is currently a theoretical concept. If (or when) achieved, solutions would have the ability to understand their own state in terms of emotions and beliefs, along with those of humans they are interacting with.