## 1 – Why Anaconda_2

A lot of times, you’ll be working with projects that require different library versions, or different versions of Python even. In my experience and many other people’s, it’s really difficult to get all these versions to play nicely together on one computer. To solve this problem, we use virtual environments, managed with libraries such as … Read more

## 1 – Why Anaconda_

A lot of times, you’ll be working with projects that require different library versions, or different versions of Python even. In my experience and many other people’s, it’s really difficult to get all these versions to play nicely together on one computer. To solve this problem, we use virtual environments, managed with libraries such as … Read more

## 5 – Search

So here’s our Sudoku puzzle. After running constraint propagation by iterating the functions eliminate and only choice we get to this point. Where in each box we’ve recorded the values that are allowed in that box. And let’s pick one at random, say this one with values eight and nine. Since we don’t know what … Read more

## 4 – Constraint Propagation

So let’s recall what we’ve done. We have two techniques to reduce our Sudoku. The first one is elimination, which consists of picking a box with a value and removing this value from all its peers. [BLANK_AUDIO] And the second one is called only choice, which consists of picking a unit, and if there’s only … Read more

## 3 – Only Choice

Now we’ll learn a technique called only choice. Let’s focus on this unit here, this three by three square. If we apply the elimination technique, we get that the only possible values for each box in this unit are these. [BLANK_AUDIO] So for example, for this box here with a 4 and a 9, we’ve … Read more

## 2 – Eliminate

Okay, here’s our first technique. It’s called Eliminate, and it works as follows. Let’s start with a simple board, and let’s say we have one box filled in. For example, this box is filled in with a seven. So what do we know? Well, we know that every row, column, or three-by-three square must contain … Read more

## 1 – Intro

Hi and welcome to the first lesson of the Udacity artificial intelligence nanodegree program. My name is Luis Serrano and I’m one of the instructors here at Udacity. Before being here I used to work in machine learning in Google. I’m the team that generated your YouTube video recommendations. In this lesson we’ll learn two … Read more

## 9 – Monty Hall Problem – Explained

To help us understand this phenomenon we’ll take the help of an old friend, probability theory. It gives us a formal mechanism to capture uncertainty in the real world. And let’s us make sensible decisions, even when we are not fully sure about something. >> Since the assignment of cars and goats is random, it … Read more

## 8 – Monty Hall Problem

It turns out that if you consistently stay with the original door, you only win about one-third or about 33 percent of the time. Whereas if you always switch, you win about two-thirds or about 66 percent of the time. So switch is indeed a better overall strategy for this game.

## 7 – Monty Hall Problem

This conversation reminds me of a game show that has a very interesting and non intuitive problem associated with it. >> Are you thinking the Monty Hall problem? >> Yes, I am. You’re shown three doors, behind one of them is a shiny new car, while the other two have a goat. These are randomly … Read more

## 6 – Tic Tac Toe_ Heuristics

What’s really interesting in this representation is, that the edges embody the rules of the game. For instance, you can only go to a finite number of next states from each current state. >> Hm, you’re right. It limits the number of moves we might need to consider, especially later in the game, when the … Read more

## 5 – Tic Tac Toe

The first option of denoting each cell as a node is fairly intuitive, assuming that we interpret the node as representing the current position of the computer player. But, does that capture enough information for you to decide what to do next? Not really. So, I guess this is not a good representation of this … Read more

## 4 – Tic Tac Toe Question

Our goal is to use a search strategy similar to the route-finding problem, but to design an AI that can play Tic Tac Toe. What, in your opinion, should be the nodes and edges in the graph that we search? Option A, should each cell be a separate node, with an edge connecting two of … Read more

## 3 – Game Playing

Finding the shortest path on a map is a very useful ability and that heuristic is a nice trick. But it seems more like an optimization problem to me. I’m not sure if I’d consider that a sign of intelligence. >> Mm, okay, how would you characterize an intelligence system or agent? >> Well, unlike … Read more

Navigation is an excellent example of a computationally hard problem that AI can help solve more efficiently. >> What do you mean by computationally hard, problems that have a high time or space complexity? >> Yeah, precisely. Let me illustrate. I’m trying to plan a road trip from Manchester to Sheffield in the UK and … Read more

## 18 – Nanodegree Outline

If all this sounds exciting to you, you’re in the right place. This nano degree program is organized into two parts, starting with this term. >> Here in term one, we’ll cover the fundamental principles of AI, search optimization, logic, planning, probability, and so on. No matter where you go in AI, these form your … Read more

## 17 – Rational Behavior And Bounded Optimality

Okay, let’s get back to our question of what is intelligence? Given these building blocks, agent, environment, perception, action, and cognition, is there a more practical definition that we can come up with? >> Well here’s one formulation that’s actually quite widely accepted. An intelligent agent is one that takes actions to maximize its expected … Read more

## 16 – Types of AI Problems – Explanations

So if we’re playing poker, that would be partially observable, I think, because you can see your own hand, but you can’t see your opponents’ hands. >> That’s true. >> It would be stochastic, because you don’t know what cards are going to be dealt and when. It would be discrete, because there are only … Read more

## 15 – Types of AI Problems – Question

One way to classify AI problems is using properties of the environment and components of state that need to be captured. >> An environment can be fully observable like Tic-Tac-Toe where you can see the entire board. Or could be partially observable like in Battleship where you can’t see your opponents ship positions. >> An … Read more

## 14 – Perception, Action and Cognition

An agent interacts with the environment by sensing its properties. This is known as perception and by producing useful output or actions that typically change the state of the environment. The process by which an agent decides what action to take based on its perceived inputs is called cognition. Much of our discussion of AI … Read more