The AI Utility Levels Schema – Building an AI Classification

This is a sneak preview to one of the features of a project I am working on. (It’s still in beta, but ripe for the sharing.) So what you have here is an AI Utility Levels matrix. It is a companion schema to the “A” through “D” AI capability levels that I first introduced in the Mapping AI Taxonomy post. Now, there’s more to this, but for the sake of brevity I will just note that together they serve as essential ingredients for building a concise AI classification system, where an AI application can be classified as, for example, a “Level A2” application. If you read the post I linked to, you will appreciate that the alpha designation ties to the AI’s interaction (user/source) capabilities and the numeric value to functional capabilities. This approach aims to help construct and apply relevant laws to AI.







Understand Automated reasoning, machine learning; most closely tied to Level A. Behavioral advertising


Perceive, understand Automated reasoning, machine learning – deep neural nets, convolutional neural networks (CNNs), recurrent neural networks (RNNs) knowledge representation. Play rule-based games such as chess or Go; tax prep chatbot; Buy goods and services based on pre-selected criteria (price, type, legal meta data).


Perceive, understand, predict (probabilistic reasoning) Automated reasoning, natural language processing; knowledge representation, advanced machine learning, deep neural nets – CNNs, RNNs; understand context. Buy goods and services based on past and predictive patterns; perform certain fiduciary duties; personhood compatible


Perceive, understand, predict, manipulate Automated reasoning, natural language processing, knowledge representation, machine learning; understand context; Artificial General Intelligence (AGI); most closely associated with the Singularity. Negotiate and execute contracts, buy, and sell virtual property, sue, settle, collect damages; personhood compatible; personal liability compatible.