Unit 1: Introduction

Objectives

• Use a Jupyter notebook to run Python code
• Start to uncover how an image classifier works.

Step 0: Get a Kaggle account

You’ll need an account on Kaggle.

You’ll also need to verify your phone number in order to get access to GPUs, but we’ll do that in the “Image Classifier” section below.

Note: You can also do this on Google Colab; you’ll see links at the notebook sections for that. If you’re using Kaggle, though, ignore the Colab links. I don’t recommend trying to run this on your own computer at this point; even if you have a compatible GPU, getting Python to work with it can be a project.

Step 1: Jupyter Notebooks

In this section, we’ll practice working with Jupyter notebooks. You may find these references helpful:

• The Help menu in Kaggle
• Kaggle documentation

A number will appear next to each of the code cells when they have run successfully.

Note carefully the difference between Command mode and Edit mode.

I highly encourage you to get comfortable with keyboard shortcuts for the following operations:

• Switch between edit and command mode. (Enter/Return and Esc)
• Insert a cell above or below (a and b)
• Change a cell to code / Markdown (m or y in Command mode)
• Run the current cell (with or without selecting the cell below) (Ctrl-Enter or Shift-Enter)

For more keyboard shortcuts click the Command Palette button on the bottom toolbar on Kaggle (it may be hidden by a cookie consent bar!) or use a search engine.

When you’re done, save your notebook and submit it on Moodle.

Step 2: Image Classifier

In the next section, you’ll work with a basic image classifier.

Note: The first time you log into Kaggle, you’ll need to Verify your phone number to get Internet and GPU access. So:

1. On the right sidebar, find “Session Options”.
2. At the bottom of the notebook options, check for a message that reads like “Want more power? Get phone verified”. Click that link and follow the instructions.
3. Turn on Internet access.
4. Switch Accelerator to GPU (either GPU option works; you might try to benchmark the difference sometime, not now).

If you have any trouble with this,

In this section (and most future Labs), the tasks to do are inside the notebook itself. You’ll find cells labeled Task and blank code chunks usually labeled # your code here. Follow the instructions top-to-bottom, then download and submit when done.

Checklist:

• Did we fill in the table with the accuracy and loss numbers?
• Did we think about whether the predictions are a probability distribution?
• Did we try out using argmax? Can we get it to show the name of the predicted class, not just its number?
• Did we try changing something to see what effect it has on the accuracy?

To help understand the landscape of AI and also get introduced to each other, let’s survey some examples that students find interesting.

In this week’s Discussion forum, please make a two-part post:

Part 1: Introduction

Post something interesting about yourself.

Part 2: AI Examples

Suppose someone asks you: Tech companies and governments are investing hundreds of billions of dollars in AI. Is AI worth the billions or is this just a bubble? Is AI actually useful for anything in the real world?

Please post a brief (50–200 word) statement where you give an example of an application of AI that you think is useful, interesting, or important. (Try to pick an example that other students haven’t mentioned yet.) Include:

• An image of the system in action, or a link to something that describes it.
• Some evidence that the technology is, or will soon be, of practical interest. Who would pay for it?
• Describe what about its behavior merits the description “intelligent”. Referring back to the contextual reading for this week, would Turing have considered this behavior “intelligent”? Would you consider it intelligent?

Also read your peers’ posts and reply to at least one.

Where to find AI examples?

• Think of some topic outside of computing that you find important or interesting. Search for “AI” or “ML” combined with that term.
• Pick a large company and look at their “AI blog” (or sometimes “tech blog” or “engineering blog”)
• Providers of AI computing, like NVIDIA, Amazon AWS, Google Cloud, Microsoft Azure, etc. like to show off what people are doing with their tools
• Lots of YouTubers, podcasters, bloggers, news sources, etc. like to talk about AI; see Resources for some examples.
• https://PapersWithCode.com/
• Subscribe to an AI or Machine Learning newsletter, podcast, or social media account. Good ones start and stop all the time, so we can’t list them here.
• AI work shows up at many academic conferences.