CVTC AI Fluency

How AI Works

You may have heard "AI", "Artificial Intelligence" or used tools like ChatGPT or Microsoft Copilot. Although AI has been used before in things like Alexa, Netflix recommendations, and translation apps; Generative AI is the technology most often thought about as "AI" today.

Generative AI tools such as ChatGPT, Microsoft Copilot, or Google Gemini "are essentially word-predictors. Based on all those prior examples of recorded text, they have a good idea of the next logical word in any given sentence. Thus, these systems don’t actually think. They don’t even comprehend the meaning of their words, leading some scholars to compare LLMs to parrots—they can mimic speech, but don’t understand what they are saying." (Yee, ii)

Step 1) Training

A computer looks at a lot of data. Think all writing, video transcripts, books, blog posts, tweets, Reddit discussions on the Internet.

Using math on this data the computer created a Large Language Model.

Step 2) Usage

You input a prompt.

The words in the prompt are translated into numbers.

Based on its training, the model predicts which word should be shown in the output.

The cycle of predicting the next work and showing it continues until all output is shown.

Quick Check

Large Language Models, like ChatGPT, generate results the same way as a search engine, like Google.

True: Incorrect, Large Language Models generate new results each time based on prediction.
False: Correct: Large Language Models do not show results from a database but generate new output each time.

Source

Yee, Kevin; Whittington, Kirby; Doggette, Erin; and Uttich, Laurie, "ChatGPT Assignments to Use in Your Classroom Today" (2023). UCF Created OER Works. 8. https://stars.library.ucf.edu/oer/8

Next Section

Deciding When to Use AI

A concern with Generative AI is that the output can be biased and prone to errors. There can be bias in the content that was used to train the model, so any output may also contain bias. Generative AI does not "check" the output. It spits out words but does not know if the output as a whole has any errors.

Generative AI may also not know about current events or recent advancements. For example ChatGPT 3.5 was not trained on any material after September 2021. To complicate things, the model might not let you know that it does not know recent events. It will output words that it knows without regard to the date.

Some models, like Microsoft Copilot, can supplement training data with Internet searches so are more likely to reply with current events. It is important to know the knowledge cut-off for the model you use.

Although the output might appear true AI might not be the best choice for your course work.

Good Uses for AI

Brainstorming Ideas: Generate ideas for projects, papers, or presentations.
Essay Drafting: Generative AI can help create initial drafts of essays based on given topics.
Summarizing Articles: Quickly summarize long articles or research papers.
Language Translation: Translate texts for language courses or international studies.
Code Generation: Assist in writing code snippets for programming assignments.

More Examples

Generating Study Guides: Create concise study guides from textbooks or lecture notes.
Creating Outlines: Develop detailed outlines for essays or research papers.
Writing Prompts: Provide creative writing prompts for assignments.
Generating Practice Questions: Create practice questions for exams or quizzes.
Paraphrasing Texts: Help rephrase sentences to avoid plagiarism.
Generating Bibliographies: Automatically format and generate bibliographies.
Creating Flashcards: Generate flashcards for studying.
Proofreading: Identify grammatical errors and suggest corrections.
Generating Visual Aids: Create diagrams or charts for presentations.
Simulating Conversations: Practice language skills by simulating conversations.
Creating Quizzes: Generate quizzes for self-assessment.
Summarizing Lectures: Summarize lecture notes into key points.
Generating Research Questions: Help formulate research questions for papers.
Creating Presentation Slides: Generate content for presentation slides.
Writing Emails: Draft professional emails for communication with professors or peers.

Bad Uses for AI

Current events: Models often have a cut-off date and are not aware of current events, even if they claim to.
List Sources: Models may make-up sources or combine sources from multiple locations.
Mathematical Calculations: Generative AI can sometimes make errors in complex calculations. Specialized calculators or software ensure accuracy and reliability.
Taking Notes in Class: AI might miss context or nuances in lectures. Note-taking apps allow for personal annotations and organization.
Collaborative Projects: AI cannot facilitate real-time collaboration and communication. Collaboration tools provide platforms for teamwork and project management.

More Examples

Programming Debugging: AI can suggest code, but debugging requires understanding the logic and context, which is best done in an IDE.
Data Analysis: AI might not handle large datasets or complex statistical methods as effectively as dedicated statistical software, which is designed for precision and depth.
Drawing Diagrams: AI-generated diagrams may lack the precision and customization needed for detailed academic work. Graphic design tools offer more control and quality.
Conducting Experiments: AI cannot physically perform lab experiments or ensure adherence to scientific protocols. Hands-on lab work is essential for accurate results.
Reading Textbooks: Active reading and note-taking help with comprehension and retention, which AI cannot replicate.
Creating Detailed Graphics: AI-generated graphics may not meet the high standards required for academic presentations. Graphic design software provides better tools for quality output.
Learning New Languages: Language learning involves practice and interaction, which AI cannot fully provide. Language apps offer structured lessons and feedback.
Managing References: AI might not format references correctly according to specific academic styles. Reference management tools ensure accuracy and consistency.
Conducting Surveys: AI cannot design or distribute surveys effectively. Survey tools offer features for creating, distributing, and analyzing surveys.
Time Management: AI might not integrate well with personal schedules and preferences. Time management apps offer customizable features and reminders.
Creating Music: AI-generated music may lack the creativity and personal touch of human composition. Music production software allows for detailed and creative control.
Video Editing: AI might not provide the precision and customization needed for high-quality video editing. Video editing software offers advanced tools and effects.
Physical Exercise Plans: AI cannot tailor exercise plans to individual needs and preferences. Fitness apps provide personalized plans and tracking.
Mental Health Support: AI cannot offer the empathy and understanding required for mental health support. Professional help and mental health apps provide better resources.
Financial Planning: AI might not account for personal financial goals and situations. Financial planning tools offer detailed and personalized advice.
Cooking Recipes: AI-generated recipes might not consider dietary preferences or ingredient availability. Cooking apps offer a wide range of recipes with user reviews.
Travel Planning: AI might not provide comprehensive travel itineraries. Travel planning apps offer detailed information and booking options.

These lists were generated using Microsoft Copilot using this prompt and some items were added, removed, or edited.

Remember if using AI, always check the output and cite your source.

Quick Check

Next Section

Effective Prompt Engineering

Typing a question or instructions to large language models is called prompting. There are two big differences between Internet searches and interacting with generative AI.

Prompts should be written in natural language using complete sentences. The more detailed the prompt, the better the output. This is much different than using key words in a Google search.
Generative AI is a conversation instead of a one-time search. You can expand on your first prompt by asking follow-up questions, clarifying your question, or asking the model to show the output in a different format. This is much different than trying different keywords when looking up information in a Google search.

Consider these prompts if you are asking a large language model about a topic that is new to you.

Good

Prompt: What are semiconductors?

Microsoft Copilot Output

Better

Prompt: Explain semiconductors and how they are used in everyday life.

Microsoft Copilot Output

Best

Prompt: Provide a two-paragraph summary of why superconductors are important. Then list the three most common semiconductors and where they are used. Explain to someone with limited knowledge about semiconductors.

Microsoft Copilot Output

Give it a try

There are several Generative AI models. CVTC recommends logging in with your credentials and using Microsoft Copilot at https://copilot.microsoft.com.

Here is a video demonstrating the use of Copilot.

ChatGPT is also an option. No account is needed to chat. https://chat.openai.com

Google Gemini can be used but an account is needed to chat. https://gemini.google.com/

Quick Check

Instructions:

You will be presented with five tasks that can be accomplished with Generative AI.
For each task, you will be shown two different prompts. Although both would give an output, one would likely give a better output.
Click or tap the prompt that would, most likely, give a better response from the model.
For each set, click or tap on the prompt that would likely give the best response.
Click or tap the next arrow to show the next set of prompts.
Click or tap the Retry button if you want to try the questions again or the Show Solution button to see the answers.

Next Section

Evaluate AI Output

Generative AI cannot let you know if the output is accurate. Remember, at a basic level it is only predicting the next word. There is not human thought or critical evaluation of the output, only a string of words. It is up to you to figure out if the output is accurate, no small task.

Consider this output from Microsoft Copilot. Click or tap each checkmark and read the details.

Consider this output from ChatGPT using the same prompt. Notice there are no sources listed. It would still be important to double-check the output to see if these are indeed facts about CVTC.

Next Section

Add the Human Element

Figuring out a good prompt, having a quality conversation, and critically evaluating the output are tasks that require humans. It is widely believed that many jobs in the future will involve humans interacting with AI. To be prepared, learning to use AI tools will be important. AI can analyze and synthesize a large amount of data and automates repetitive tasks. This can allow humans to focus on creativity, critical thinking, and complex decision-making. The human element can bring context, empathy, experiences, and ethical considerations to AI systems.

Quick Check

Instructions:

You will be presented with three sets of items.
For each set, drag one to Human if the task would be better accomplished by us or drag to Generative AI if the task would be better accomplished by AI.
Repeat for the other sets of items.
Click or tap the Submit button to check your answers.
Click or tap the Retry button if you want to try the questions again.

Next Section

Protect Your Work

It is also a good practice to document your process for creating work. Saving notes, Internet search history, Generative AI chat history, and drafts of work can be used to help demonstrate the work is your own if questions arise. Here are a few things to consider when completing course work.

Tools that auto-complete as you type use AI to predict the next word. This writing may be flagged as AI generated since the tools look at the next word.

Turn off Text Predictions in Word.

Click Text Predictions: On from the bottom status bar.

Uncheck the box before Show text predictions while typing.

Click the OK button.

Turn off Smart Compose in Google Docs.

Click Tools > Preferences.

Uncheck the box before Show Smart Compose suggestions.

Click the OK button.

Tools that check your grammar use patterns to determine sentence structure. This writing may be flagged as AI generated since the tools look for similar patterns in writing.

Check your online Microsoft Word or Google Docs using TurnItIn Draft Coach.

Next Section

Citing Sources

Whenever using Generative AI to help with academic work, it is important to give credit. This is similar to citing a source from a book or Internet source.

Check out the CVTC library Citation Styles Guide for help with citations.

Click the menu items for APA, MLA, Chicago, and Artificial Intelligence.

Artificial Intelligence (AI) Statements

The standard CVTC syllabus and Student Handbook include a statement on AI. If you have questions on these statements, be sure to ask your instructor.

Syllabus Statement

Individuals may use AI (Artificial Intelligence) tools (such as ChatGPT, generative AI) to aid in learning and academic endeavors. It is essential to acknowledge the role of AI as a support tool and maintain academic integrity by ensuring that all submitted work is original and properly cited. Any act of academic dishonesty, including but not limited to plagiarism or cheating, will follow the College policy on academic dishonesty found in the student handbook.

Handbook Statement

At CVTC, we believe that emerging technologies, including AI, have the potential to transform teaching and learning, and we are committed to exploring and adopting innovative technologies that enhance student success and aid learning.

We recognize that emerging technologies are constantly evolving, and we are committed to staying current with the latest trends and developments to provide our students with the best possible educational experience. Therefore, technology should be used to enhance, rather than replace, the traditional learning experience.

We believe that technology can be a powerful support tool for promoting student success for all, and we are committed to ensuring that all students have access to the technology they need to succeed.

Ultimately, our goal is to leverage emerging technologies in ways that promote student success, foster innovation and creativity, and prepare our students for the rapidly changing world they will encounter after graduation.

How AI Works

How AI Works

Step 1) Training

Step 2) Usage

Quick Check

Source

Deciding When to Use AI

Deciding When to Use AI

Good Uses for AI

More Examples

Bad Uses for AI

More Examples

Copilot Prompt

Quick Check

Effective Prompt Engineering

Effective Prompt Engineering

Good

Better

Best

Give it a try

Quick Check

Evaluate AI Output

Evaluate AI Output

Add the Human Element

Add the Human Element

Quick Check

Protect Your Work

Protect Your Work

Citing Sources

Citing Sources

Artificial Intelligence (AI) Statements

Syllabus Statement

Handbook Statement

Quick Check