Get to Know AI: The Basics

September 18, 2024

Welcome

Today

1. What is AI & Generative AI (GenAI)?
2. How Does it work?
3. What are some of the concerns with AI?
4. How do I use it effectively?

AI Prompt: create an image of an auditorium filled with eager participants in a workshop on generative ai

“Digital Literacy in AI”

Presenters

Roberto Cásarez, Ph.D.

Academic Technology and AI Literacy Specialist

Office of Information Technology

Brandon Rich

Director, AI Enablement

Office of Information Technology

Ardea Caviggiola Russo, Ph.D.

Director, Office of Academic Standards

Office of the Provost

What is AI?

Basic AI Terms

• AI (Artificial Intelligence) - ability for technology to accomplish complex goals typically associated with human intelligence
• Machine Learning - subset of AI where computers are trained to learn from data and improve over time without being explicitly programmed for every task
• Deep Learning - a type of machine learning that mimics how the human brain works using artificial neural networks
• Narrow AI (Weak AI) - AI systems designed to perform a specific task
• General AI (Strong AI) - a future concept where AI systems would have the ability to perform any intellectual task that a human can do

Prompt: Create a friendly image of an AI

Brief History of AI

1950 - Alan Turing proposes the idea of a machine that can "think"

1956 – The term "Artificial Intelligence" is coined by John McCarthy at the Dartmouth Conference

1966 – The first AI chatbot, ELIZA, is developed by Joseph Weizenbaum

1997 – IBM’s Deep Blue defeats world chess champion Garry Kasparov

2000s – AI advances with the rise of machine learning algorithms and larger datasets

2016 – AlphaGo, developed by DeepMind, defeats a world champion Go player

2020s - Generative AI models like GPT-3 and ChatGPT demonstrate AI’s ability to generate human-like text

Examples of AI

• Healthcare
• Medical Diagnosis, Personalized Treatment, Drug Discovery
• Retail
• Product Recommendations, Inventory Management, Customer Assistance
• Education
• Personalized Learning, Virtual Tutors, Grading and Assessment
• Security
• Facial Recognition, Cybersecurity, Smart Home Devices

What is Generative AI (GenAI)?

• AI is a field of study, much like statistics
• Generative AI a subset of AI
• It is a set of algorithms that can create/generate seemingly new, realistic content -- such as text, images, and audio -- from a set of training data.
• GenAI, as we know it now, came into commercial use in 2014
• Use exploded after OpenAI released ChatGPT on November 2022

How Does It Work?

Types of Generative AI

• Text-to-text
• ChatGPT, Claude, Perplexity, Gemini,
• Text-to-image
• Dall-e, Midjourney, Flux
• Text-to-audio
• Whisper, ElevenLabs, Udio (music)
• Text-to-video
• Sora, Runway

12

Sora AI Prompt: The camera follows behind a white vintage SUV with a black roof rack as it speeds up a steep dirt road surrounded by pine trees on a steep mountain slope, dust kicks up from it’s tires, the sunlight shines on the SUV as it speeds along the dirt road, casting a warm glow over the scene. The dirt road curves gently into the distance, with no other cars or vehicles in sight. The trees on either side of the road are redwoods, with patches of greenery scattered throughout. The car is seen from the rear following the curve with ease, making it seem as if it is on a rugged drive through the rugged terrain. The dirt road itself is surrounded by steep hills and mountains, with a clear blue sky above with wispy clouds.

Data and Training

Large language models are static text engines with a knowledge cutoff date.

Classroom Assistants

Faster Reimbursement

Transfer Articulation

Personalized Student Chatbot

Web Accessibility

Language Model Capabilities

Baseline features / use cases

• summarizing large documents
• analyzing images and data
• writing a position description
• writing code
• preparing for an interview
• analyzing contracts

Bolted-on features

• searching the web
• executing their own code
• making calls out to third party services
• revising work based on conversation

Training vs Context

Permanent knowledge "baked in" to the model

Temporary knowledge retained only for a session:

• your chat history
• your uploads

Training vs Context

Permanent knowledge "baked in" to the model

Temporary knowledge retained only for a session:

• your chat history
• your uploads

data privacy

depends

on terms!

AI Models vs Chatbots

• AI "models" such as gpt-4o are the "engine"
• Chatbots like ChatGPT or other products are like the "car"
• Sometimes they have similar names 🤷‍♂️

24

ChatGPT prompt: "please make a graphic in a flat shaded cartoony, but professional style showing a car at 3/4 perspective with its hood open and engine visible"

Language Models in the GenAI Landscape

OpenAI – most powerful, most features. Models include gpt-4o and gpt-o1.

​

Microsoft – CoPilot products use OpenAI's gpt models.

​

Anthropic. OpenAI competitor's "Claude" models are nearly as powerful as gpt-4o.

Google – "Gemini" models

​

​

Perplexity offers search-like chat with citations using gpt-3.5

​

​

Meta offers the free and open "llama" models

​

​

New models incorporate planning

Generated with flux.ai. Prompt: photo-realistic image of a person solving a newspaper crossword puzzle. shallow depth of field with focus on the person and the puzzle. brightly-lit background could be a living room or a kitchen. you see the person from behind; the puzzle is clearly visible.

Ethan Mollick

Wharton School

What should we be concerned about?

Bias and Fairness

• AI systems are trained on data produced by people who have conscious or unconscious biases (e.g., the internet)
• This can cause the AI to encode and amplify these existing prejudices, since they will appear in its output
• It isn’t hard to imagine how this could lead to harmful outcomes in criminal justice, healthcare, hiring, lending….

Economic Impact

• Job displacement due to automation, which will disproportionately affect lower-wage jobs
• Access to high-quality AI tools may come at a higher cost, and profits from the use of these tools may advantage company owners rather than workers
• “AI colonialism”
• Dependence on AI systems may increase vulnerability to failures and errors (remember the CrowdStrike outage in July?)

Privacy and Surveillance

• AI requires lots of data, which likely means increased collection and increased risk of breaches
• AI-powered surveillance systems can track individuals without their knowledge; potential for abuse by bad actors
• Gen AI can create deepfakes, which can be used to spread misinformation or damage reputations
• The ability of AI to correlate data can be used to infer sensitive information about individuals that they never intended to disclose

​

Environmental Concerns

• AI models require significant computing power, leading to increased electricity consumption
• Data centers contribute to carbon emissions, and cooling systems require large amounts of water
• Manufacturing of AI hardware requires elements that impact the mining industry
• Electronic waste from discarded AI hardware poses disposal challenges, especially those that contain toxic metals
• AI companies may choose to prioritize short-term efficiency over long-term environmental risks

How Do I Use AI Effectively?

Data use at ND

• Non-sensitive data: Avoid personal, confidential, or proprietary information.
• Anonymized data: Data stripped of personally identifiable information.
• Aggregated data: Collective data that masks individual contributions.
• Public datasets: Openly available data from reputable sources.

​

Ethical and Responsible Use

• Use AI in alignment with the university's mission and values.
• Respect intellectual property and data privacy rights.
• Be transparent about the use of AI.
• Strive for fairness and avoid bias.
• Consider the potential societal and environmental impacts of AI.

General Advice

• You are responsible for the output and use of AI
• ChatGPT will not get reprimanded or in trouble for what you submit
• You need to check the information
• Do not rely solely on AI; combine it with your own critical thinking and knowledge.
• Use AI where you believe it can make an impact
• Seek opportunities and support to learn how to use AI tools effectively and ethically
• Cite or acknowledge AI when called for

AI Tools

• ChatGPT
• Google Gemini
• Microsoft Copilot
• Claude
• Perplexity
• Dall-e (images)

Prompting

• Prompting is the process of giving specific instructions or input to an artificial intelligence system in order to generate a desired response or output.
• The quality of the response often depends on how the prompt is structured
• Garbage-in, Garbage-out
• Use it as an assistant, and interact as if it was
• Move beyond the Google Search

Prompting Formula

Jeff Su (Director). (2023, August 1). Master the Perfect ChatGPT Prompt Formula (in just 8 minutes)! [Video recording]. https://www.youtube.com/watch?v=jC4v5AS4RIM

Mandatory

Important

Nice-to-Have

Examples

• Bad Prompt
• I don't want to attend the meeting
• OK Prompt
• Write and email that states that I cannot attend the meeting.
• Good Prompt
• You are the manager of a team of administrative staff overloaded with work on the “Get to Know AI” Project. You have been asked to attend a meeting by your supervisor to discuss the project details, but would lose valuable time working on the project if you attend the meeting. Write a sincere email stating that you cannot attend the meeting. Provide some options in the email to address any questions at a future time, or via a different method.

Closing

Reflection and Resources

• Key Points
• Understand the basics of generative AI and how it works
• Be aware of possible risks of AI
• Get equipped with practical knowledge of how to use AI effectively
• Resources on AI.ND.EDU
• Join us in Future “Get to Know AI” Workshops

Questions

Disclosure: The majority of the images in this presentation were AI generated

We will be around if you have additional questions

Student/Faculty Questions

Technology/Faculty Questions

Technology/Staff Questions

AI In-practice/Any

Questions