AI Classification – An Update to the AI Application Categories

I updated the AI Application Categories table as part of a project that I am working on, so consider this a sneak preview. The reference to “CP” is to what I refer as “AI core principles,” a list (currently) comprising 35 design and policy foundations for AI, which will be released as part of my ongoing work. Note that the first version of this classification approach was released in 2012. For more on that, you can read here.

Level

Capabilities

Permit Requirement

CP Adherence Level

A

Programmatically limited to interact with internal input sources (the end-user) and is incapable of operational variance. (Dictation, visual object recognition.)

None

Low

B

Respond to end-user queries and commands and can interact with internal and external data sources (e.g., the web). Siri, Alexa

None

Low

C

Autonomous and dynamically selects from internal or external sources and decide on optimal ways to interact with the end-user.

Optional

Elevated

D

Autonomous, can reprogram behavior, replicate, recursive self-improvements, imitate, and exceed human behavior. Super-intelligent AI.

Mandatory

High