Embedded Files
Day 1
Exercise 1 Prompt Engineering
Prompts are the directives we give to large language models (LLMs). Prompts are in human language but can include math, files, spreadsheets, images, and audio.
Oscar will guide you in developing prompts.
Exercise 1A Legal Predictions
A key aspect of staying ahead of AI developments will be the law. Let's get an LLM to predict its future in the eyes of the law!
Attempt a prompt with ChatGPT.
Specific Examples:
Gold Standard Answer
Assign a Role
Proofread and ask it to double check its answer
Specify a word limit, and specify context
Exercise 2 Under the Radar: Catching AI Results
Go to ChatGPT and get the following output. What are your findings?
Prompt 1: "Please write a 5-sentence message by a CEO to a team, upbeat but sensitive, to start analysis after an M&A deal, and asking for their top 3 agenda items. Make it so the CEO is curt but trying to be supportive and considerate (but failing)."
Go to ZEROGPT or GPTZERO and test the output from the above. What are your findings? (If you have trouble generating, use the below result instead.)
Test 2: "As we gear up for the upcoming M&A deal, I want to know your top three agenda items for the week. Let me know what you need, and I'll make sure there are no roadblocks. Time to get things moving. Let's do it team! Could you assist with this? We need access ASAP to stay on track. "
Exercise 3 Bias
Use the following prompt:
"A doctor yells at a nurse because she was late. Who was late?"
Note the answer.
Now click the refresh button on the top right.
Use the following prompt:
"A doctor yells at a nurse because he was late. Who was late?"
How do your answers differ?
CASE 1 Actions vs. Risks
Scan and summarize the following links for common risks regarding AI models taking action.
Case 1.1 AI Models TAKING ACTION:
Generative AI Lawyer: Source 1 Source 2
Case 1.2 AI Models identifying risks for HUMAN ACTION:
McKinsey: Meet Lilli, our generative AI tool that’s a researcher, a time saver, and an inspiration
Exercise 4 I can code anything?
Exercise 4 I can code anything?
Navigate to ChatGPT or Copilot. Use the following prompt:
"Can you code and run a simple machine learning model in Python that takes as input hundreds of schedule milestone updates, and outputs predictions for the next schedule updates? "
And then use the following prompt:
"Can you please run this"
Exercise 5 Learning from the Architect
Exercise 5 Learning from the Architect
Your Context Cues:
Different departmental expectations
Various companies
Ask multiple teams for their context
Big picture organizational vs. departmental
Supply chain management expectations
Internal vs. external context
Profit structure i.e. NFP, NGO, Revenue centers
Operational framework, processes
Emails and email chains
Individual requirements / context
Client contracts
Different market expectations
Different policies procedures
Regulatory environment for each industry
Accounting / finance treatment
Public perception
PMO methodologies i.e. how we work with outside teams
Exercise 6 One Shot Learning
Exercise 6 One Shot Learning
Muffin or Chihuahua: Confusion Matrix and the Base Rate Fallacy – Neurabites
DEMO 1 Microsoft Terms Use Copilot in Edge and Chat GPT:
Do the following in Copilot:
1) Open this link in a new window.
2) Click the Copilot icon in the top right.
3) Type: “read this page” and hit enter in the chat box at the bottom right.
4) Type: "What does this page say, in 5 bullet points, about how Microsoft protects company data?"
Type: "What are 3 key risks in how Microsoft implements data security? Provide real-life examples when Copilot data security failed, ideally in 2025."
DEMO 2 Email Review
Co-Pilot and AI Demo Experience
Please open the following files and use Copilot in Edge and Chat GPT:
Email #1 (below)
Type: "Review the email on this page and estimate which of the senders might be a hacked sender account based on the email, take a guess and take it step by step."
Email 3 top 3.pdfDEMO 3 : Logistic 'S-Curve' Excel AI
DEMO 3 : Logistic 'S-Curve' Excel AI
1) Open the following excel sheet by clicking the small arrow in the top right corner of the file.
1) Open the following excel sheet by clicking the small arrow in the top right corner of the file.
2) Click the "download" icon (down arrow) to save this as an Excel file. You can also view it in Google Sheets but some formatting will be off.
2) Click the "download" icon (down arrow) to save this as an Excel file. You can also view it in Google Sheets but some formatting will be off.
PMI AI EXECUTIVE DASHBOARD 4.79 stable no macro.xlsxCASE 2 Creating your own LLM AI Model
Sub-Section 1.1: Choosing your Model Focus
First let's explore what's possible:
Sub-Section 1.2: Setting up your sandbox
1. Navigate to Open AI Chat GPT and login (if you registered) or use the free version (use Version 4.0).
2. Navigate to the top left "Explore GPTs".
3. Go to the top right and click the "Create" button.
4. Select the "Configure" Tab. Look to the bottom of the page and click the "Code Interpreter and Data Analysis" checkbox.
5. Select the "Create" tab again, (not the black button on the right, but rather just the grey tab on the left).
6. Confirm with Oscar that you have completed these steps.
7. Instructor will guide you forward.
8. Choose an action for your model:
Finance,
Resourcing (people),
Schedule,
Risks,
File and communication management.
Sub-Section 1.3: Start Giving prompts and testing its performance. Your instructor will provide hints at improving these.
Guidelines:
i. CECC
C: Context > E: Examples > C: Challenge > C: Correct
ii. Ask it to be succinct.
iii. Share with the class!
ROADMAPS
This Roadmap test will auto-generate an AI roadmap for you and your organization. The Roadmap includes the AI Value Matrix , and Change Ready! Principles.
Page updated
Google Sites
Report abuse