AI is about making better, faster, cheaper decisions

The utility of all analytics, from the most basic to the most advanced, is in helping you make decisions (Decisions matter)1.

It isn’t about seeing into the future. While many seem to treat it like a crystal ball, the most you will get to with that approach might a magic eight-ball. AI is more of a robot, than an oracle.

How do I decide which items to recommend to someone? How do I decide if this image is a cat or a dog? How do I decide what the right price of this would be?

Noticing that AI2 is about decisions helps us make choices about what class of methods to deploy against problems that arise.

The Cynefin framework classifies the contexts of decisions into obvious, chaotic, complicated, and complex. In chaotic contexts, just do something and learn. In obvious contexts, use rule-based systems. In complicated contexts, use AI to learn the rules, to automate the automation. In complex contexts, use AI as sensors.

  1. I use decisions broadly – as the intent behind actions. In this sense it is a superset of predictions and not limited to the kind that are exposed in decision-support systems. All ML tasks can be framed as decisions. ↩

  2. I use AI broadly – as intelligence residing outside humans, encompassing the whole stack of analytical tools available. Slightly incorrect, but useful and trendy. ↩

Notes mentioning this note

There are no notes linking to this note.

Here are all the notes in this garden, along with their links, visualized as a graph. If you don't see any nodes try zooming and panning in the grey area.