CIM563 Design with AI

AI System

A general term describing a computer program that has an AI algorithm to process input and generate output.

Ultra-intelligent Machine

A machine that far surpasses all the intellectual activities of man, however clever.

Image Recognition

A recognizer can recognize what an object is, as well as where in the picture it is.

Algorithm

A sequence of steps to solve problems

Machine Learning

A subfield of AI that focuses on creating algorithms for AI systems to learn from examples or experiences.

Mental Model

A user's concept of how an object should work based on its appearance.

Agent

Anything that can be viewed as perceiving its environment through sensors and acting upon that environment through actuators

Types of agentive technologies

Assistive, automatic, agentive

General AI

Can perform various tasks with human-like flexible skills

Design Signifier

Communication devices to let users know what an object is capable of. Affordances are implicit, signifiers are explicit (signs, pictures)

Rule-based AI algorithms

Create AI systems by first figuring out the knowledge that human experts have, and translating that knowledge into rules that the system can apply to a task.

Acting rationally

Doing the right thing.

Narrow AI

Focuses on a specific task

Data-driven AI algorithms

Gives AI systems both correct and wrong answers, or set up an environment in which the system can experiment on its own to gain experiences (learn).

Precision

How good an ML model is at predicting a specific category. True Positives / All Positives or True Dogs / All dogs, real + false positive

Recall

How many times an ML model was able to detect a specific category. Positives / Relevant Positives or Dogs Predicted as positive / All dogs in the set

Agentive Lightbulb

Identifies your activity, and sets lighting appropriately.

Assistive Lightbulb

Switch - manually or by voice - on/off. Show the electricity consumption of each bulb.

Accuracy

Tells you how many times an ML model was correct overall.

Intelligence

The ability to acquire and apply knowledge and skills | The ability to learn or understand to deal with new or trying situations

Strong AI

The agent can actually "think"

Weak AI

The agent can possibly act intelligently, but it only simulates the action without really "thinking"

Data Mining / Data Analytics

The application of ML methods to large databases

Thinking humanly

The automation of activities we associate with human thinking, such as decision making, problem solving, learning, etc.

Image Classification

The output is given as one of a predefined class

Training a machine

The process of using algorithms to build models for decision making

Design Affordance

The relationship between an object and a person; provides answers as to what the function of an object is, typically through visual appearence.

Acting humanly

The study how to make computers do things which, at the moment, people are better.

Thinking rationally

The study of mental faculties through the use of computational models.

Non-technical definition of AI

The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision making, and translation between languages.

Reinforcement Learning

Trial-and-error based. The algorithms figures out what to do based on a reward/penalty system. ("Learning How to Swing")

Automatic Lightbulb

Turn the light on at 6PM, off at 11PM

Supervised learning

Uses prepared, labeled data

Unsupervised learning

Uses unlabeled data. The machine clusters information based on likeness.