Instructor: Alan Koch

Foundations of Artificial Intelligence (ICP-FAI)

Posted on April 29, 2024 by -

1. Introduction to Artificial Intelligence

Evolution of AI

Understanding AI

  • Key concepts
  • Potential applications
  • Limitations of AI in the workplace
  • Practical applications and potential of AI
  • How AI differs from traditional computing

Pieces of the AI Puzzle

  • Machine learning
  • Deep learning
  • Algorithms
  • Data processing
  • Generative AI

Evolving State of AI

  • Why is AI evolving at such a rapid rate?
  • Artificial Narrow Intelligence (ANI) vs. Artificial General Intelligence (AGI)
  • Common myths and misconceptions surrounding AI

Exercise – (In the style of Jeopardy)

  • Identify concepts
  • Define terms
  • Which enabled the other?

2. Ethical and Legal Considerations of AI

Ethics in the Context of AI

  • Moral principles and guidelines
  • Ethical considerations to address

The Inherent Bias of AI

  • The impact of bias
  • Approaches and strategies to ensure ethical AI
  • Human validation of AI

Data Compliance and Privacy

  • Regulatory considerations (e.g. GDPR & HIPAA)
  • Compliance strategies

3. Prompt Engineering

Introduction to Prompt Engineering

  • The basics of effective AI Prompt Engineering
  • Recognizing well-crafted prompts that lead to useful, specific answers

Effective Prompts

  • Techniques for crafting prompts that produce accurate, reliable, and unbiased results
  • Recognizing when to revise prompts based on the AI's results
  • The importance and impact of context in prompt engineering
  • Creating prompts based on clearly identified goals

All Prompts are Not Created Equal

  • Contrast results from various LLMs and Generative AI tools
  • Sources for syntax guidance for different AI solutions
  • Patterns and strategies for crafting effective prompts Exercise
  • Use several prompting techniques with ChatGPT

4. Artificial Intelligence in the Enterprise

The Agile Advantage

Agile Mindset and AI

  • The connection between the agile mindset, values, and principles and AI
  • How a culture of learning, reflecting and adapting enables AI success

Agile Behaviors and AI

  • Applying iterative development, continuous feedback, and collaboration to AI solutions
  • How planning, designing, testing, and keeping AI systems up to date differs from other efforts.

Cross-Functional Teams and AI

  • The cross-functional skills needed in a team focused on AI (e.g., prompt engineering, data literacy, data science, software engineering, ethics, domain-specific knowledge)
  • Evolving the cross-functional AI team

5. Leveraging AI in the Organization

Business Value of AI

  • How AI can be used to create competitive advantages, optimize operations, and enhance customer engagement
  • Making strategic decisions about investing in, developing, or implementing AI solutions based on AI's business value
  • How AI supports and complements the skills and creativity of humans

Align AI with Strategy

  • Aligning AI initiatives with strategic objectives
  • Infrastructure to support AI initiatives
  • The cultural shift towards data-driven decision-making
  • Oversight for responsible AI governance

AI Initiative in the Real World

  • Examples of real-world AI initiatives
  • Types of AI solutions and their potential impacts on an organization

Exercise – Plan your Path Forward

  • What steps can you take to use AI right away?
  • What will you plan to do in the next few months?
  • What will take longer to achieve, but would be worth it?

ChatGPT Prompt Engineering Primer

Posted on May 30, 2023 by -

Part 1: ChatGPT Basics & Underlying Concepts

We begin by spending just a little time describing ChatGPT and defining relevant terminology to set the stage for learning. The Q&A allows learners to quickly confirm their understandings so we can get to the fun and interesting part, using this tool.

  1. Definitions
  2. Types of GPTs (Chat & others)
  3. The role of data (and your company’s data)
  4. What ChatGPT is good at (and not good at)
  5. Who can use ChatGPT (and for what)?
  6. Q&A

Part 2: Interacting with ChatGPT

Most of our time in this short course is spent interacting with ChatGPT. Each of the sub-sections below introduces a specific usage mode, and includes:

  • Examples – The instructor shows multiple examples of that usage mode.
  • Play-time – Learners are allowed time to try their hand with that usage mode on their own.
  • Show and Tell – Learners share and discuss their experiences with that usage mode with each other and get feedback from the instructor.

Converse with ChatGPT

We start with the most basic interactions with ChatGPT:

  • Ask questions
  • Use chained prompts (and break chains when needed)
  • Use strong and weak verbs in prompts.
  • With Examples, Play-time, Show and Tell

Prompt ChatGPT to tailor its response

ChatGPT’s responses will sometimes not meet your needs. They may be too long and verbose, too short and cryptic, or not laid out in a useful way. So, we will look at various ways to ask it to provide or restate its response in more appropriate ways.

  • Summarize long responses
  • Expand on short responses
  • Format responses (e.g., bulleted or numbered lists, headings, etc.)
  • Provide an example of content & format you want it to respond with

Prompt ChatGPT to use appropriate roles

We will look at how to focus ChatGPT’s interactions with you based on role definitions.

  • Provide responses appropriate for a particular role or audience
  • Provide responses appropriate to a channel (e.g. social media)
  • Play a role in how it responds
  • Role-play with you

Prompting ChatGPT to respond with questions

ChatGPT can do more than just answer questions. We will look at how to prompt it to ask questions as well.

  • Prompt it to provide questions (e.g. for an interview)
  • Prompt it to ask clarifying questions before responding

Providing data to ChatGPT

The free public version of ChatGPT cannot access data. So, we will look at ways you can provide it with the data you want it to use in its responses.

  • Loading the prompt with data
  • Providing data in multiple prompts for a single response
  • With Examples, Play-time, Show and Tell

Part 3: Advanced Use Cases & Interactions

Connecting ChatGPT to other applications

We will explore a variety of tools that can be used to connect applications to the paid version of ChatGPT.

  • Plug-ins for commercial applications (e.g. Excel) to use ChatGPT
  • ChatGPT’s API for writing your own plug-ins or adding ChatGPT capabilities to your company’s applications

Making data available to ChatGPT

We will explore a variety of ways to enable ChatGPT to use other data.

  • Plug-ins for ChatGPT to access external data sources (e.g. Wikipedia)
  • Tools to feed large data sets into ChatGPT’s prompts
  • Using a private LLM (Large Language Model) to enable ChatGPT to use specialized language and terminology
  • Grounding ChatGPT on your company’s data

Part 4: Class wrap-up and Q&A

Effective User Acceptance Testing

Posted on October 17, 2020 by -

Part 1: What UAT Is and Is Not

  • We will begin by describing how UAT differs from other software testing and how it fits into various software development lifecycles (including Waterfall and Agile). Along the way, we'll define a variety of key terms and identify the players.

Part 2: Understanding the Business Need

  • Business Need has many dimensions from correct computations to ease of use. We will explore each of those dimensions so you can ensure that your UAT addresses each of them in an appropriate way.

Part 3: What Could Go Wrong?

  • Of all the things we could test, which should we focus on? We will apply Risk-Based testing to focus our UAT where it will be most valuable.

Part 4: "U" is for User

  • Effective UAT includes testing from the standpoint of all of the users (both active and passive ones). We will discuss ways to identify all of the users and ensure that their viewpoints are included in our UAT.

Part 5: Incremental UAT

  • UAT is usually the final gate before deployment, but any problems found at that point in the project can be costly and time-consuming to correct. So we will introduce an incremental approach to UAT that can be integrated into any software development lifecycle (even Waterfall). This incremental approach enables you to identify issues earlier (when they are easier to fix), and reduces the likelihood of unpleasant surprises at the project's end.

Part 6: Preparing Test Data

  • Good tests require appropriate test data. We will discuss how to identify and prepare test data that will enable good Acceptance Testing. Along the way we will discuss the limitations, dangers and (in some cases) illegality of using production data for testing, and we will look at options for addressing those issues.

Part 7: The Acceptance Test Plan

  • As the old adage says, "Fail to plan; plan to fail." The plans for UAT will be different from those for other testing activities. We will provide guidance for UAT plans, including how to find the "sweet spot" of providing enough guidance to ensure effective and repeatable tests, while enabling the testers to exercise the system as they will use it after it is deployed.

Part 8: Performing UAT

  • Testing is more than just using a computer. We will provide guidance for how Acceptance Testers should operate while performing UAT. We will discuss following UAT plans as well as going beyond them to explore how the system works. We will also discuss evaluating test results, reporting issues and raising questions.

Part 9: "A" is For Acceptance

  • We will finish with a discussion of deciding if the system is acceptable or not. We will explore this both from each tester's perspective, and for UAT as a whole. Along the way, we will talk about "minor" defects, unresolved issues, and what it means for the system to be "good enough" in a particular context.