Nomenclature for Levels of Artificial Intelligence (AI)

Jack Krupansky
4 min readJun 26, 2023

--

People use the term AI or artificial intelligence in a lot of confusing ways so that we can never be sure what they are talking about. This informal paper (note) attempts to lay out a somewhat more comprehensible nomenclature for the term AI.

I think of AI as a number of levels of intensity of function and integration of intelligence capabilities:

  1. Automation or Classical automation. Just about anything you can do with a computer that a human can do. All the basic stuff. Hardly worth calling it AI, but it does require human intelligence and it is artificial.
  2. Advanced automation or Advanced classical automation. Verges on what we now consider AI, but people were reluctant to call it AI to avoid the stigma that was traditionally associated with AI.
  3. Weak AI or Narrow AI. Relatively simple forms of AI. Single functions. Single modules of AI. Corresponding to single human mental functions. It is common to see AI classified as Weak AI vs. Strong AI.
  4. Moderate AI. Somewhat more sophisticated AI. Integrating multiple functions. Multiple, integrated modules. Corresponding to multiple human mental functions.
  5. Strong AI or Broad AI. Much more sophisticated AI. Integrating many functions. Many integrated modules. Corresponding to many human mental functions. But generally still short of full human-level intelligence (AGI — artificial general intelligence), although some people might interpret the two terms as synonyms.
  6. AGI or Artificial General Intelligence. Implements full human-level intelligence. A very wide range of modules, fully integrated. Corresponds to virtually all human mental functions.
  7. ASI or Artificial Superintelligence or Superintelligence. Achieves a level of intelligence significantly greater than human-level intelligence. Integrated modules that achieve a greater level of intelligence than the comparable human mental functions. May implement capabilities beyond those of human mental functions, either simply more capable functions or even additional functions. Sometimes a synonym for Kurzweil’s Singularity, but doesn’t have to achieve more than some significant increment above human intelligence.
  8. Kurzweil’s Singularity. A superintelligence that grows even greater over time, learning on its own, at an exponential rate. Far surpassing anything human intelligence can hope to achieve. Integrates biology as well as computers, genetics, nanotechnology, robotics and artificial intelligence.

So, the next time you encounter a reference to AI or artificial intelligence, try to assess which level the writer or speaker is referring to. There’s a good chance that it won’t be entirely clear.

So what is ChatGPT, Bard, Bing Chat, et al? I would say that ChatGPT, et al are somewhere between Moderate AI and Strong AI. Proponents would say it’s definitely Strong AI, while critics would say that it’s not quite there — my term Moderate AI might be more appropriate, at least for now. But nobody can with a straight face make a strong claim that AGI — Artificial General Intelligence — has been achieved, yet.

Animal intelligence?

For the purposes of this level of discourse, animal and human intelligence are integrated. Some levels of animal intelligence can in fact be significantly more difficult to implement than some levels of human intelligence, such as sensory perception and motor control.

Also, a lot of animal intelligence can be more related to robotics — movement and interacting with real-world objects.

But a fair fraction of the intelligence of humans is shared with animals, just not the higher-order human-only intellectual mental functions.

Robotics?

Much of the intelligence of robots is either animal-level intelligence, such as sensory perception, motor control, movement, and interacting with real-world objects, or human or near human-level intelligence covered by the nomenclature above.

There probably needs to be a separate nomenclature for levels of mobility and interaction with real-world objects than the nomenclature for intelligence given in this informal paper.

But to the extent that a robot is a thinking machine, that thinking is covered by the nomenclature of this informal paper.

Robots and AI systems

The concepts of AI can be implemented in two main forms:

  1. AI systems. A computer system in which AI is used to implement functions as a computer application or networked web service. Produces informational results or media output, or sends commands to other computer applications or networked web services.
  2. Robots. An electromechanical device capable of physical mobility and/or the ability to sense its physical environment, gather data from and about its physical environment, make decisions with a significant degree of autonomy, and perform activities and/or have physical effects in its physical environment, again with a significant degree of autonomy. Typically using AI, but the AI capabilities are used internally to support the overall robotic activity rather than the purely informational results or media output or computer commands of a non-robotic AI system.

Autonomy and agency?

Autonomy and agency are two entire dimensions of AI which are not covered by the nomenclature of this informal paper. Autonomy refers to the degree of freedom that an intelligent entity has to choose how to act in any situation. Agency refers to the degree to which an intelligent entity is obligated to act relative to the goals and constraints set for it by another intelligent entity, its principal, for which it is an agent.

For more detail on autonomy and agency, see my informal paper:

And for deeper detail, see my informal paper:

References

For more detail on AI, see my informal paper:

And for a lot more depth, see my informal paper:

For more of my writing on AI: List of My Artificial Intelligence (AI) Papers.

--

--