Getting Started with

Generative AI as a
Teaching Partner

Dale Pike
Associate Vice-Provost for
Technology-enhanced Learning
Virginia Tech

https://creativecommons.org/licenses/by-nc/4.0/
Welcome/Introductions
Objectives
Understand simplified Explore basic AI Experiment with
fundamentals of chatbot use cases for prompting strategies
generative AI teaching & learning

Explore the basic concepts Investigate the fundamental Engage in practical

and principles of generative use cases of AI chatbots in experimentation with
AI in a simplified manner. educational settings for different prompting
teaching and learning strategies to understand
purposes. their impact on generative AI.
Quick Poll PollEv.com/dalepike606
Artificial Intelligence

Regona, Massimo & Yigitcanlar, Tan & Xia, Bo & Li, R.Y.M. (2022). https://doi.org/10.3390/joitmc8010045
Generative AI
Create New Content
Text, images, music, audio and video.

Powered by foundation models

Can multi-task and perform out-of-the-box tasks, including
summarization, Q&A, classification, etc.

Adaptable
With additional training, able to refine performance for specific use
cases.
Beware the Hype
Cautious
Optimism
Why should we be cautious?
Bias Privacy Costs (environmental,
social, financial)

Economic Academic Hype

disruption integrity
Why should we be optimistic?

Productivity gains

Potential for personalized learning

Potential for learning analytics

Regona, Massimo & Yigitcanlar, Tan & Xia, Bo & Li, R.Y.M. (2022). Opportunities and adoption challenges of AI in the construction industry: A PRISMA review. Journal of Open
Innovation Technology Market and Complexity, 8(45). https://doi.org/10.3390/joitmc8010045
Finance Writing Assistance
● Fraud detection on credit cards ● Autocorrect (Microsoft Word, Google Docs)
● Personalized insurance rates ● Grammar checkers
● Robo-advisors for investments ● Plagiarism detectors
● Loan approval algorithms ● Autocomplete/predictive text
Security ● Sentiment analysis
● Facial recognition (FaceID, airport security) Communication
● Cybersecurity threat detection ● Spam filters for email
● Scored trustworthiness of online identities ● Smart reply suggestions in email/text messaging
Dating Apps ● Virtual assistants (Alexa, Siri)
● Suggested matches ● Chatbots for customer service
● Profile assists Entertainment & Shopping
Health & Wellness ● Netflix recommendations
● Step counters and activity trackers ● Amazon recommendations
● Menstrual cycle prediction ● Spotify playlists and Daily Mixes
● Calorie estimation for food logs ● Audible recommendations
● Early disease diagnosis assistance ● Video game matchmaking
Photography Navigation & Transportation
● Automatic photo categorization ● Google Maps/Waze for driving directions and
● Object/scene recognition traffic updates
● Facial recognition for tagging ● Uber/Lyft price estimates and ETAs
● Photographic style transfer ● Autonomous vehicles
What is Generative AI?
Generative AI or generative artificial intelligence refers to the use of
AI to create new content, like text, images, music, audio, and
videos.

Generative AI is powered by foundation models (large AI models)

that can multi-task and perform out-of-the-box tasks, including
summarization, Q&A, classification, and more. Plus, with minimal
training required, foundation models can be adapted for targeted
use cases with very little example data.

Sources:
https://cloud.google.com/use-cases/generative-ai
Large Language Models
D

P
Generalized capabilities
Fine Tuning
What’s the Big Deal?

Human-like

General-purpose public use

Omni-use workflow
https://www.mckinsey.com/featured-insights/mckinsey-explainers/whats-
the-future-of-generative-ai-an-early-view-in-15-charts#/
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4573321 https://www.oneusefulthing.org/p/centaurs-and-cyborgs-on-the-jagged
How is it being used?
Company LLM Application Future LLM

OpenAI [(GPT-3.5), GPT-4] (ChatGPT), [ChatGPT+] GPT-5?

Microsoft GPT-3.5, GPT-4 Copilot (Bing) GPT-5?

Google PaLM 2, Gemini Pro Gemini Gemini Ultra

Anthropic Claude 2 Claude Claude 3?

Meta Llama 2, Code Llama Hugging Face, GitHub Llama 3?

MIDJOURNEY

Vegan Buddha Bowl, Colorful selection woodblock print, cardinal in summer GQ cover photo of Abraham Lincoln,
of fresh vegetables and grains, Zen-like stylish modern white shirt, short hair,
simplicity, Serene and Balanced, Natural handsome, fashion Photography, studio
and Minimalist styling, Soft diffused lighting
lighting
MIDJOURNEY

“College student caught in the rain while walking across campus.”

Setup
1 Microsoft Copilot for Bing 2 Blank doc for scratch notes
(or ChatGPT or Claude 2) (text editor, Google Doc,
etc.)
Utilize tools like Microsoft Copilot, Prepare a blank document for
ChatGPT, or Claude 2 for Bing to taking scratch notes using a text
enhance the workshop editor or Google Doc to capture
experience. key insights during the workshop.
Context window:
The context window refers to the range of tokens or words the AI model can access and consider when generating
responses to prompts or user inputs.

The context window functions as the dynamic memory of a conversation, containing the AI's responses and the user's
prompts.

The context window size for GPT-3 is 2K (~1,500 words) and for GPT-4 it is 8K, or ~10k words.

Anthropic Claude 2 has a context window of 100k tokens, or ~75k words.

Omni-use
What do prompts do?

● Prompts can provide context - They give the AI system

information about what kind of response is expected or relevant to
the conversation. This makes the system's responses more
specific and appropriate.
● Prompts can control tone/style - The wording and content of
prompts impact the tone, personality and style of the AI's
responses.
● Prompts can encode preferences - Carefully crafted prompts
can encourage the system to respond helpfully, harmlessly, etc.
R.A.C.E.

Roles (I am/This is for/You are)

Action Specific description of desired activity

Internal (within prompt) or external (general knowledge,

Constraints uploaded reference, etc.) boundaries to guide focus of response

Examples Specific demonstration of response format, tone, structure,

etc.

Useful to help focus thinking

I am an instructor planning a university Intro to Film course. This is for my
students, to engage them in learning about film history and critique. You are a tool
to help generate educational content.
Create a detailed prompt for a course module focusing on the evolution of film
techniques from the silent era to modern digital filmmaking, including key
milestones.
Use a postmodern lens to frame the prompt,
exploring how films reference and reinterpret other films and media.
The module could start with the Lumière brothers and end with contemporary
digital effects, highlighting innovations by directors like D.W. Griffith, Alfred
Hitchcock, and George Lucas.
Practical use cases

Extract info Draft a form Tutor Meta

from a list or report (Feynman prompting
Technique)
Practice

https://bit.ly/genai-handout
Questions
Using AI for Instruction
• Focus on faculty use cases
• “Advanced beta” (note inaccuracies, perception, etc.)
• Stage your exploration
AI as course design assistant

R
“Propose a completely new way to teach the concept of sustainable
development using virtual reality (VR), including an outline of the VR
experience and how it can lead to deeper understanding and
engagement. Include potential pre- and post-VR activities that integrate
with ChatGPT for reflection and discussion.”

M
“Design an interactive, scenario-based learning activity on climate change
impact, where students can choose different paths and see the consequences
of their choices in real-time.”

A "Create a series of quiz questions based on the key topics of the Environmental
Science syllabus and provide explanations for the correct answers."

S "Generate a list of key topics for a course syllabus on Environmental Science."

 SAMR
Substitution
Augmentation
Modification
Redefinition
 SAMR
Substitution: AI-Enhanced Material Preparation
Generative AI Use: Automating lecture notes and slides creation
Human-Centric Focus: Professors personalize AI-generated content
 SAMR
Augmentation: AI-Integrated Lecture Enhancement
Generative AI Use: Enriching lectures with current examples and
research.
Human-Centric Focus: Professors curate AI-suggested materials for
relevance.
 SAMR
Modification: Interactive and Dynamic Learning Experiences
Generative AI Use: Creating real-time quizzes, simulations, or virtual
environments
Human-Centric Focus: Professors guide interactive learning,
emphasizing critical engagement
 SAMR
Redefinition: Personalized Learning Pathways
Generative AI Use: Tailoring follow-up activities and content to
individual students
Human-Centric Focus: Professors mentor and integrate
personalized paths with course objectives
Process over product

Avoid high stakes summative Offer practice with

assessments feedback

Frequent low-stakes formative Mastery

assessments Learning
Case Study
Case Study
Bloom’s Two Sigma Problem

https://web.mit.edu/5.95/readings/bloom-two-sigma.pdf
What happens?
Tutoring
How are the interactions with a learner different during
tutoring?

Mastery Learning
What are the core principles of mastery learning?
Mastery Learning
1. Baseline, or diagnostic testing;

2. Clear learning objectives, sequenced as units usually in increasing difficulty;

3. Engagement in educational activities (e.g., deliberate skills practice, calculations, data

interpretation, reading) focused on reaching the objectives;

4. A set minimum passing standard (e.g., test score) for each educational unit;

5. Formative testing to gauge unit completion at a preset minimum passing standard for mastery;

6. Advancement to the next educational unit given measured achievement at or above the

mastery standard; and

7. Continued practice or study on an educational unit until the mastery standard is reached.

https://tea.dtei.uci.edu/resources/mastery-learning/
Responsible AI

● Validity and Reliability

● Safety

● Security and Resiliency

● Accountability and Transparency

● Explainability and Interpretability

● Privacy

● Fairness with Mitigation of Harmful Bias

NIST Trustworthy and Responsible AI standards Image created using DALL-E with prompt: “a beautiful
https://www.nist.gov/trustworthy-and-responsible-ai non-textual visualization of the concepts of Privacy
and Security”
Resources

One Useful Thing Ethan Mollick explores generative AI in higher education

AI Snake Oil Sayash Kapoor and Arvind Narayanan put a critical lens on AI hype

7 Things You Should Know About Generative AI | EDUCAUSE Review

2023 EDUCAUSE Horizon Action Plan: Generative AI

AI Preparedness Checklist | 1EdTech

TRUTH
TRUST