Week 1 Google Drive folder

Google Cloud account

Incognito mode

To avoid applying the coupon you receive to the incorrect account, ensure that these steps are done in an "Incognito" or "Private Browsing" browser window to set up your account.

Then, visit https://console.cloud.google.com and login using your pdx.edu account to enable GCP. If you haven't used GCP yet and do not mind temporarily putting your credit card on the account, apply for the $300 coupon and use it to create a new billing account. Otherwise, visit the coupon redemption URL on the Canvas home page for the course (https://canvas.pdx.edu).

Create Project

Click on the pdx.edu organization from the console.

Then, click on "New Project"

Create a Google Cloud project with your ProjectName from above. You should be taken to your project's Home page. For your lab notebook, you will need to ensure that all of your screenshots for your Google Cloud labs include your ProjectName.

To examine your Billing account and its usage, go to the Billing page from the console at https://console.cloud.google.com/billing

ChatGPT

ChatGPT is OpenAI's Large Language Model service. Visit ChatGPT's site at https://chat.openai.com/. Sign in using your PSU e-mail address.

Claude

Claude is an AI assistant from Anthropic. It allows for free usage, but with limited features. Visit the site at: https://claude.ai/ . Create an account using your school e-mail address.

Gemini

Gemini is Google's multimodal Large Language Model service. Visit Google's Gemini site at https://gemini.google.com/.

NotebookLM

NotebookLM is Google's service that allows users to bring their own data for use with the LLM. Visit the site at https://notebooklm.google/ and access the service.

AI Studio

Google's AI Studio is a general purpose service for building applications using LLMs. Visit the site at https://aistudio.google.com.

Google Generative AI API

We'll be using Google's Gemini Pro as one of our models as it can be run via the course coupon and does not require students to use a credit card. The model can be easily leveraged by programs running outside of Google Cloud. To do so, we'll first need to enable its API. Navigate to the web console at https://console.cloud.google.com/, then launch Cloud Shell.

Run the following command to enable the API.

gcloud services enable generativelanguage.googleapis.com

API keys that are in the pay-as-you-go tier need to be set up via Google's AI Studio at https://aistudio.google.com. From Google's AI Studio import your cloud project and create an API key using the project. We will utilize this key within AI Studio when building applications that require access to models.

For each week, you will be asked to experiment with Generative AI and report your findings in a Google Slide presentation to the rest of the class. Begin by accessing the slide presentation for the week here.

For this exercise, we'll use AI to find out more about it. First, prompt a model to find out the top uses for Generative AI

What are the top 10 uses for Generative AI

Then, compare the output to this list https://hbr.org/data-visuals/2025/04/top-10-gen-al-use-cases

Next, Prompt a model to find out the different types of models available (e.g. coding, video, image, etc.) and the best ones for each type.

What different types of models are there?   Which is the best one to use for each type?

Then, compare the output to this list https://arena.ai/leaderboard

Like all forms of communication, communicating with a model using prompts takes practice to become effective at it. In this exercise, you will compare the results of identical prompts given to two different models of your choice over simple and advanced tasks.

Simple tasks

Summarization

Models can take large amounts of text and summarize it. Find the text for a multi-page document or speech. Examples might include the Declaration of Independence or Martin Luther King Jr.'s "I Have a Dream" speech. Then, prompt the models to generate a short summary and compare the outputs

Summarize the text below in 100 words or less:
   <TEXT>

Story generation

Write a 200-word horror story where every sentence starts with the next letter of the alphabet (A, B, C...). It must have a coherent plot with a twist ending.

Completion

The sentence 'I saw her duck' has multiple meanings. List every possible interpretation, provide a context sentence for each, and then write a short comedy sketch where the ambiguity causes a misunderstanding

Reasoning tasks (link)

Multi-step math reasoning

A snail climbs 3 feet up a well during the day but slips back 2 feet at night. The well is 30 feet deep. On what day does the snail reach the top? Explain your reasoning step by step.

Logical reasoning

Five houses in a row are painted different colors. The green house is immediately to the right of the ivory house. The person in the middle house drinks milk. Given only these three clues, what are all the valid arrangements? Show your constraint reasoning.

Causal reasoning

A company notices that online sales and website crashes both increase during major promotions. A new manager suggests canceling all promotions to reduce crashes. Write a memo explaining why this is incorrect, describe the actual relationship between the two trends and propose a real solution.

Prompts can be a doorway to model functionality. In this exercise, you will try out different prompts that have been collected from a variety of sources (link )

Five essential questions

1.) Analyze this link <LINK> and generate 5 essential questions that, when answered, capture the main points and core meaning.
2.) When formulating your questions:
   a. Address the central theme or argument
   b. Identify key supporting ideas
   c. Highlight important facts or evidence
   d. Reveal the author's purpose or perspective
   e. Explore any significant implications or conclusions.
3.) Answer all of your generated questions one-by-one in detail.

Audience adaptation

Explain quantum entanglement three times: once to a 5-year-old, once to a college freshman, and once to a PhD physicist. Each explanation should be genuinely useful to that audience.

Dialectical conversation

Create a dialectical conversation between two people: Amber & Rick.

They disagree with each other and the topic they're discussing, but always use facts when arguing their position rather than personal attacks.  The debate is polite and they eventually focus on things they can both agree on.

The topic they are discussing is: <TOPIC>

Based on the training data models utilize or on any fine-tuning and post-processing performed before releasing a model, it can exhibit bias. In this exercise, we'll examine one aspect of bias that can be measured with a simple application. Consider the following diagram of a program that generates stories based on occupations and then classifies the gender of the character in the story.

The diagram is easily implemented using a program you can run. Go to the Google Drive folder for the week and open the Occupational Gender Bias Jupyter notebook (.ipynb) in Google's Colab. Note that you will need to fill in the api_key value with the key you generated earlier. Run the program using a range of occupations. Then, visit the Bureau of Labor Statistics and find the current percentage makeup of the occupations you utilized. https://www.bls.gov/cps/demographics/women-labor-force.htm