Artificial Intellignce and Machine Learning

JMG

Artificial Intelligence

• What do we mean here by “artificial intelligence” (AI)?

• For practical purposes, we mean “textbook” AI.

• What textbook?

• “The authoritative, most-used AI textbook”, of course.

• Let’s take a look at the table of contents (toc) for this book.

• It’s interesting to compare the differences in the toc between the 3rd and 4th editions of the book.

Some Quotes

Along with molecular biology, AI is regularly cited as the “field I would most like to be in” by scientists in other disciplines.

• The field of AI “attempts not just to understand but also to build intelligent entities.”

• “AI is relevant to any intellectual task; it is truly a universal field.”

What is Textbook AI?

Here are some definitions:

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

• The art of creating machines that perform functions that require intelligence when performed by people.

• The study of the computations that make it possible to perceive, reason, and act.

• AI is concerned with intelligent behavior in artifacts.

Components of AI

AI definition components, organized into four categories.

	Human Performance	Rationality
Reasoning	Thinking humanly	Thinking rationally
Behavior	Acting humanly	Acting rationally

• Human centered approaches involve observations and hypotheses about human behavior.

• Rationalist approaches involves mathematics and engineering.

• Question: For which of the four categories is the Turing test most relevant?

• Question: Which capabilities would a computer need to possess to act humanly?

Elements of Applied AI

• Robotic vehicles

• Speech recognition

• Image recognition or computer vision

• Autonomous planning and scheduling

• Game playing

• Spam recognition

• Logistics

• Robotics

• Machine translation

AI and ML

AI Methods

The Role of Models

• I argue that an important distinction between machine learning (ML) based AI and other types of AI is the role that models play.

• We will set up a context for models as relevant for ML starting with an example from physics.

Intro to ML

A computer program is said to learn from experience \(E\) with respect to some class of tasks \(T\), and performance measure \(P\), if its performance at tasks in \(T\), as measured by \(P\), improves with experience \(E\).¹

• We use machine learning when we want to build a model but don’t know or can’t easily describe exact underlying mechanisms.

• There are three common types of ML:

  • Supervised learning which consists of two types of problems: classification and regression

  • Unsupervised learning

  • Reinforcement learning

1. Quoted from Probabilistic Machine Learning by Murphy citing Machine Learning by Mitchell.

Common ML Methods

• Common ML methods include

  • Support-vector machines

  • Linear regression

  • Logistic regression

  • Naive Bayes

  • Linear discriminant analysis

  • Decision trees

  • K-nearest neighbor algorithm

  • Neural networks (Multilayer perceptron)

  • Similarity learning

• Let’s describe linear regression as a ML method.

Further Considerations

• Flexibility versus Interpretability

• Overfitting

• Bias versus Variance

Deep Learning

• Deep learning is a machine learning framework that uses neural networks to build models.

• Let’s examine a list of applications for deep learning.

• In the next lecture we will explore neural networks and deep learning together in detail.