Problem Solving with Artificial Intelligence

Thinking about building an application using the concepts of agents and environments provides a way to create any artificial intelligence application. This section provides an overview of the basic concepts that underlie every artificial intelligence system.

We review some of the common assumptions about the actions of agents, including what they know and how they make decisions. We find that combining simple agents together in a larger system can bring us to true intelligence - in both biological and technological systems.

In this section we introduce Builder, one of the first robotic artificial intelligences created. Following this, we review the relationship between how living things act in their environment and how internal states can lead to sophisticated decision-making processes that solve problems.

We introduce the concept of search, a way to contemplate potential future actions, within an intelligent system.

We first start with uninformed search: strategies for decision-making with no additional information.

We wrap up this module talking about some of the practical considerations of the methods that we have reviewed in the course so far.

We start with best-first search and discuss why it is more effective compared to other search methods we’ve seen so far.

In this section we take a deep dive into Dijkstra’s algorithm to build our experience implementing an intelligent strategy to solve a simple problem.

We review two of the most important search algorithms, A* and AO*.

We expand our view of heuristics and include machine learning as a way to learn a heuristic from data.

Shakey the Robot is one of the most interesting artificial intelligences created. We introduce Shakey and its environment and learn how this robot solved problems using the techniques in this section and how many of these techniques are still used today in self-driving cars and in the Mars Rover, Curiosity.

We begin by looking at the theory of competition and formalize what we mean by strategy, gameplay and winning a game.

We take a closer look at games with perfect information, such as chess or tic-tac-toe, and games with imperfect information such as poker. How can we create a strategy in each of these cases?

Games can become very complicated and cumbersome to analyse. In this section we discuss alpha-beta pruning - a technique to simplify our analysis.

Chance introduces another consideration to playing games. We discuss optimal strategy in games that include a chance element such as backgammon.

Finally, we review some of the practical considerations of games and creating artificial intelligence to play them.

We start by exploring ways to set up constraints that can eliminate part of our search space.

Now that we have our constraints defined, we’ll review how to use them to solve problems like the map coloring problem.

We start by looking at the kinds of problems where the path to the solution doesn’t matter and introduce local search.

Local optima can cause many different problems. We investigate ways to avoid local optima including simulated annealing and local beam search.

Evolutionary algorithms are based on ideas in biological science. We review ways that these algorithms can be applied to problems in artificial intelligence.

What does it mean to solve a game? We take a deep dive into checkers, one of the first problems solved in part by machine learning.

In artificial intelligence, chess has always been a major field of study. In this section, we review some of the historical approaches to AI chess including Mechanical Turk, El Ajedrecista and Deep Blue.

Games with imperfect information like poker are notoriously challenging. In this section we review the technology that resulted in the poker-playing AI’s Cepheus and Deepstack.

Robotics and Artificial Intelligence is closely linked. We review some of the most interesting robotics projects in AI.

Artificial intelligence is an extremely important topic in the business world. This section covers several important applications including the Autonomous Vehicles, Scheduling, and Speech Recognition.

This final section sums up the concepts we’ve covered in the course with a short send-off.