Chapter 2. Data, Data, Data

Maybe if I know where it all came from, and why, I would know where it’s all headed, and why.


Data is the fuel that powers most AI systems. In this chapter, we will understand how data, and devising methods for extracting useful and actionable information from data, is at the heart of perception AI.

Perception AI is based on statistical learning from data, where an AI agent, or a machine, perceives data from its environment, then detects patterns within this data, allowing it to draw conclusions and/or make decisions.

Perception AI is different from the three other types of AI:

Understanding AI

Where an AI system understands that the image it classified as a chair serves the function of sitting, the image it classified as cancer means that the person is sick and needs further medical attention, or the textbook it read about linear algebra can be used to extract useful information from data.

Control AI

This has to do with controlling the physical parts of the AI agent in order to navigate spaces, open doors, serve coffee, etc. Robotics have made significant progress in this area. We need to augment robots with “brains” that include perception AI and understanding AI, and connect those to the control AI. Ideally, like humans, control AI then learns from its physical interactions with its environment by passing that information to its perception and understanding systems, which in turn pass control commands to the agent’s control systems.

Awareness ...

Get Essential Math for AI now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.