1 of 15

Exploring Generative AI in Pharmacy Education

Created using “ChatGPT for PowerPoint”�*not created by ChatGPT

May 9^th, 2024

Dr. Ralph Lugo

Dan VanZant

2 of 15

Today’s Agenda

Part 1: Background and Overview

Part 2: Demo

Part 3: Real-World Use Cases

Part 4: Wrap- Up: Q&A, Discussion and Play

In today’s workshop Dr. Lugo and I are going to introduce you to Generative AI tools available to faculty, staff and students, some pitfalls and things to consider when using them in your day today and in the classroom and finally we’ll show you some real world examples of how you can use these tools in your day to day.

Note: We are by no means experts on these tools. This field of AI tools is rapidly evolving with new versions and new players entering the market every day. The list of tools we are going to mention and explore today is by no means exhaustive. Also, the list of abilities that we are going to show you is by no means exhaustive. Our goal with this workshop is to help you to understand a little of what these tools can do and help you be more informed, more comfortable using them in your own work as administrators, faculty, and students.

At the end of this Lunch and Learn you should

Possible Learning Objectives

Understand the Fundamentals: Describe the basic functionality and underlying technology of generative AI models like ChatGPT.
Recognize Applications in Pharmacy Education: Identify potential applications of AI tools in enhancing teaching, research, and administrative processes within the pharmacy curriculum.
Evaluate Ethical Considerations: Discuss the ethical implications, including privacy, bias, and reliability, of using AI in academic settings.
Explore Hands-On Use: Demonstrate how to interact with ChatGPT to generate educational content, create exam questions, or simulate patient interactions.
Assess Limitations and Accuracy: Evaluate the limitations of AI tools in terms of accuracy and scope of knowledge, especially in specialized fields like pharmacy.
Plan for Integration: Develop strategies for integrating AI tools into daily tasks and long-term projects, considering both the technological needs and the potential resistance or enthusiasm of colleagues.
Stay Informed on Developments: Provide resources for ongoing learning and updates in the field of AI as it evolves.

Possible Learning Outcomes

Describe AI Capabilities: Participants will be able to articulate what generative AI models like ChatGPT can and cannot do within the context of academic and professional pharmacy settings.
Name examples of Generative AI tools they have access to: Attendees will be able to identify and locate existing AI tools available to them.
Identify Applications: Faculty and staff will identify at least two practical applications of AI tools that could enhance their teaching methods, research projects, or administrative workflows.
Recognize Ethical Implications: Participants will be able to discuss key ethical issues related to the use of AI in education, such as data privacy, consent, and potential biases in AI-generated content.
Demonstrate Proficiency in Basic Use: After hands-on exercises, participants will demonstrate the ability to interact with AI tools to perform specific tasks relevant to their roles, such as generating educational materials or designing patient scenarios.
Critically Evaluate AI Content: Participants will critically assess AI-generated content for accuracy and relevance, especially in the context of evidence-based pharmacy practice.
Commit to Continued Learning: Faculty and staff will express a commitment to keeping abreast of developments in AI technology as it pertains to their field, noting specific resources or avenues for ongoing education.

3 of 15

Introduction and Overview

4 of 15

Different Types of �AI Tools

Chat Bots of Old

Smart Devices

Generative Pre-Trained Transformer (GPT) Launched in November, 2022

Media Algorithms

Tools with AI integration –fitness trackers, healthcare apps, cameras, digital assistants and of course smart phones

Media Algorithms - social media and streaming media service algorithms and smart assistants.

Chat Bots of yesteryear - 24hr Microsoft Tay – designed to learn from interactions with Twitter users, IBM’s Watson. These older chatbots were more rigid; their responses were based on much smaller data sets and used more structured response rules – more like decision trees or referencing a database

Generative

Pre-Trained on the WWW. - something in the neighborhood of 175 billion parameters

Transformer refers to the response architecture designed to handle data sequences and consider the context of each word, phrase, or block of text are more complex - and Pre-Trained using 175 billion parameters. This allows them to use context Context, Generative,

Generative AI refers to AI models that generate content such as text, images, and videos. It is not limited to simple tasks and can produce complex, creative outputs. However, it's important to clarify the boundaries and capabilities of Generative AI to avoid misconceptions and exaggerated expectations.

GPT

GPT 2

GPT 3

GPT 4

5 of 15

Terms and Definitions

Tokens

Specific GPT designed for natural language interactions using a chat type interface

Photos provided by Pexels

ChatGPT

Large Learning Model (LLM)

Prompts

Natural Language Processing (NLP)

Hallucinations

AI trained using vast amounts of data

Questions, instructions, or statements input into chatbots

combination of computational linguistics and machine learning

Illogical or factual inconsistencies generated by chatbots

Units of text such as letters or words used to measure processing quantity.

Bias

Skewed responses based on the limitations of the training data.

6 of 15

Terms and Definitions

Internal values are placed on aspects of training data such as patterns, context clues, vocabulary, syntax, etc.

Photo generated by DALL-E

Parameters

Training Data

API

Data fed into an LLM

Application Programming Interface allow two or more applications to communicate with one another.

7 of 15

What can Generative AI generate?

Text responses
Formatted Text
Office Documents
Images
Graphics
Audio
Video
Code
Websites

8 of 15

Popular Generative AI Tools

ChatGPT by OpenAI
CoPilot by GitHub/Microsoft
Claude by Anthrop\c
Gemini by Google
Meta AI by Meta (Facebook)
DALL-E by OpenAI
Lex by Amazon

9 of 15

Additional AI Assisted Tools

Microsoft Office
Google Suite
Adobe Creative Cloud
Grammarly
Many, if not all of the major web-based tools have launched or are working on adding generative AI based features.

10 of 15

Model	Company	Release Date	Max Tokens	Key Features	Pricing	Platforms
ChatGPT 3.5	OpenAI	Nov, 2022	Not specified	Unlimited messages, interactions, and history Training data through Sept. 2021	Free	iOS, Android, Windows and Mac Browsers
ChatGPT-4	OpenAI	March, 2023	4096 tokens	Data from 2023 and before. Can use DALL-E, browsing the internet, create and use custom GPTS. Can upload and generate documents. Fewer hallucinations. Safer than 3.5. Can accept images as an input. Can handle 25,000 words of text. Training data through	$20/month
CoPilot	GitHub / (Microsoft)	Nov, 2023	Not specified	Code suggestions, integration with dev environments	Free�$20/month
CoPilot �Protected	“”	Dec, 2023		Free with ETSU Office 365. Does not store store or use your chats to train the AI . Does not allow you to upload documents.	Free for ETSU
Claude 2.1	Anthrop\c	March, 2023	200K tokens	Reduced hallucination, tool use, large context handling, allows users to upload documents in the free tier	Free �$20/month
Gemini�(Bard)	Google	March, 2023	Not specified	Supports creative applications, model chaining	Usage-based pricing; specific rates depend on usage
Amazon Lex	Amazon	April, 2017	Not specified	Conversational interfaces for apps, voice and text	Pricing based on requests; details on AWS pricing page	No direct interface
Llama 3	Meta	April, 2024	Not specified	Multilingual models, large context windows	Not publicly commercialized; used internally and in research contexts	No direct interface
DALL-E	OpenAI	Sept, 2022	Not applicable	Image generation from textual descriptions	Usage-based pricing; specific rates depend on usage	Multiple
Meta AI	Meta Platforms	Sept. 2023	Not specified	Various AI projects, research-focused	Free in Meta products	Meta Tool Specific

11 of 15

Let’s Interact with ChatGPT

12 of 15

Interactive Demo

Dan Demonstrates a basic chat interaction

Dan will show you a basic chat interaction.

We’ll form groups of 3-4 and use a specific prompt.

Photos provided by Pexels

Team Exercise with a specific Prompt

13 of 15

Prompt for Group Interaction

Give me the basic steps I could use to teach someone how to take a a blood pressure reading.

14 of 15

Considerations and Caution

FERPA and HIPAA
Hallucinations
Bias
Plagiarism
Shift in workload
Over-Reliance

15 of 15

Wrap-up and Discussion

Q&A, Discuss, and Experiment

Photos provided by Pexels