OpenAI recently provided updated information about their upcoming GPT store launch. The short of it, wait until next year. They sent this update via email presumably to everyone who has started using the Custom GPT creation tool they launch in November.
We are continuing to make improvements to GPTs based on your feedback. To improve Actions we updated the configuration interface, enabled one click testing, added debug messages in preview, and now allow multiple domains. There have also been questions around uploaded files. Uploaded files are downloadable when using Code Interpreter so we’ve made this feature default off and added messaging to better explain this. If you have additional feedback, we’d love to hear from you here.
I have to admit I didn’t take the time to try to figure out Actions because they are tied to APIs which I don’t fully understand. I’m assuming if you know what you’re doing you can provide additional functionality to your GPT using existing tools and API keys. I am interesting in diving deeper into using my own files as data sources to instruct my GPT how to provide very specific instructions for faculty.
One of the downsides I’ve found with sharing conversations from ChatGPT is you can’t share chats that include images. At least not yet.
ChatGPT also doesn’t offer any easy method for managing shared links. Once you’ve shared a conversation, you have to return to the original chat which can be hard to track down especially if you’re a heavy user. Looking quickly at my chat history, I have at least six new chats a day with some days easily tripling that number.
With these limitations in mind, I was happy to see how the recently updated Bard handles shared chats. Previously the share icon that appears at the bottom of each chat in Bard only allowed you to send a conversation using gmail, or save it as a Google Doc. Now you also have a share option that allows you to quickly create a public link.
Bard will recommend a headline based on the prompt you are sharing, but it also includes a generate more button to create new headline options. In the image below the original headline was What is this? which is the prompt I used when I uploaded the graphic.
I clicked on the Generate More button and got several better options to use for my shared link. You could also click on the pencil icon to provide your own Headline for the public link.
You have the option to share an entire chat or just the prompt you selected.
With the preferred Headline selected just click on Create Public Link and you get immediate confirmation that the link was created.
From this screen you can easily post the link to your favorite social media channels, or copy the link for use anywhere else you’d like. Here is the Public Link I just created. The image below shows what this page will look like for anyone who has the link. There are no privacy settings, so the link could be shared broadly.
I chose this particular chat to demonstrate that Bard has no problem sharing conversations with embedded images. Once Bard is capable of image generation, this will be very beneficial.
The second point I wanted to make here is that Bard gives me more control over managing all the public links I’ve created. Click on the settings icon (the gear shown here) and you get a few options in Bard including access to your public links.
From this screen I can easily access and manage any links I’ve shared. Deleting a public link does not remove it from your Bard chat history. And, it looks like these links automatically expire in 6 months. I’m not sure how I feel about that, but otherwise I appreciate this level of access and control over all the links I’ve shared.
I knew you could do a lot to control the output of ChatGPT. I’ve used it to create multi-column tables and to generate cut-and-paste code, and I guess I knew it could spit out a spreadsheet. But, until recently I did not realize that it could also output a word document.
I was testing an AI Teaching Consultation GPT I’ve been building. Using this tool I was able to have ChatGPT generate a Syllabus Template,
There does seem to be a couple significant limitations though. For one, when you share a conversation with a document embedded, recipients can’t download that file.
The second issue is that the original link also expires after a short time. When I went back to check the link a second time I got this notice.
So, there are possibilities here, but I am betting that Bard, with it’s tie in to the Google suite, does a better job of creating sharable documents.
A month or so back I was exploring the ability to use ChatGPT as a mini application by building a customized menu of options for next steps. This post looks at a similar process.
This tip comes from the same author and takes the earlier idea of a menu of options even further with a focus on providing a list of ‘operation contexts’ which are essentially different roles. Here is a quick example based on the prompt included in the linked article.
Creating a story about artificially intelligent teddy bears for an adult audience is a fascinating and unique challenge. To achieve this, I suggest the following operational contexts, ranked from most to least optimal:
Creative Writing Assistant: As a language model with extensive knowledge of storytelling techniques, genres, and character development, I can assist in brainstorming ideas, developing plot points, and fleshing out characters. This context is optimal because it leverages my primary strength in processing and generating text, specifically tailored to creative writing.
Interactive Writing Collaborator: In this context, I act as an interactive collaborator, where you present initial ideas or drafts, and I provide feedback, suggestions, and alternative perspectives. This context is beneficial because it combines your creative vision with my ability to offer diverse ideas, thus enriching the story development process.
Genre and Theme Consultant: Leveraging my extensive database of literary works, I can provide insights into various genres and themes that might align with your story about AI teddy bears. This context is useful for ensuring that the story resonates with adult audiences and remains unique and engaging.
Literary Research Assistant: In this role, I can assist in researching background information, historical context, or technological concepts relevant to your story. This context supports the story’s authenticity and depth, particularly if you wish to incorporate specific scientific or cultural elements.
Draft Reviewer and Editor: Finally, I can function as a reviewer and editor, offering constructive feedback on drafts of the story. This includes checking for coherence, consistency, and grammatical accuracy. This context is the least optimal compared to the others, as it is more reactive than proactive in the creative process.
Each of these contexts provides a different level of engagement and support, tailored to your specific needs as a writer and the unique aspects of your story.
I’m exploring this additional functionality of user selected roles for authoring learning objectives. I started with the operational context script and added my own instructions. I am using this approach to test one of my first custom GPT focus on writing learning objectives. Using this feature I can also upload documents for my GPT to access for additional information.
I think I’ve mentioned previously that Claude AI has the ability start a conversation around a pdf file. This means you can use existing structures, frameworks, and models to help create unique assignments for any course. I’ll use the TILT model as an example for this post. Below is an introduction of this model from the TILT website.
TILT Framework
“The following resources from the Transparency in Learning and Teaching project (TILT Higher Ed) can help faculty, educational developers and administrators to apply the Transparency Framework (of purpose/task/criteria) in contexts including assignments, curricula, assessment and strategic initiatives, all toward the goal of enhancing student success equitably.”
This TILT template is available as a pdf document through the TILT website.
So, I can give Claude this file and ask for an assignment appropriate for my course using this template. The basis of TILT is to clearly document the Purpose, Task and Criteria to provide students with a clear picture of the assignment requirements and outcomes.
I provided Claude with the following prompt:
And here is the assignment that was created.
Here is a draft assignment using the transparent assignment template for a digital color theory course:
Assignment: Color Palette Project Due date: End of Week 5
Purpose: The purpose of this assignment is to help you practice selecting, evaluating, and applying colors digitally to convey specific moods or themes. These skills are important for success in digital design and art fields. Through this project you will become more skilled at:
Understanding color theory, its terminology and methods
Analyzing how color combinations evoke different moods/emotions
Synthesizing your knowledge of color theory to curate color palettes for specific purposes
Evaluating and selecting the most effective color palettes
Knowledge: This assignment will help you become familiar with:
The basics of color theory like hue, saturation, value
Complementary, analogous, and triadic color schemes
Tools and techniques for selecting, sampling, and applying color digitally
Task: Create three unique color palettes, each containing 5 colors, using digital design software. Each palette should convey one of the following themes: energetic, soothing, mysterious. Provide a brief rationale for each chosen color and how it contributes to the overall mood of the palette.
Criteria for Success: Characteristics of excellent work include:
Colors are selected thoughtfully to create a cohesive palette
Palette colors evoke the stated mood/theme
Explanations show understanding of color theory and palette choices
Palettes and text are visually appealing and clear when presented
To achieve excellence, avoid selecting colors arbitrarily. Instead, intentionally curate colors and provide insightful rationale. Vague, minimal explanations will result in an average grade. This assignment is worth 15% of your total course grade. Reflecting on your finished, graded project will help improve future assignments.
Please let me know if you would like me to modify or expand on any part of this assignment draft. I’m happy to incorporate any feedback to better meet your course needs.
I ask Claude to iterate on this just a bit and then asked for a rubric for the assignment.
Here is the rubric updated to have the assignment worth 100 points:
Criteria
Points
Excellent
Good
Fair
Poor
Followed Instructions
10
Followed all instructions accurately
Followed most instructions
Followed some instructions
Did not follow instructions
Understanding of Color Theory
10
Showed deep understanding
Good understanding
Basic understanding
No understanding
Energetic Palette
10
Cohesive energetic palette
Mostly energetic palette
Somewhat energetic palette
Not energetic palette
Soothing Palette
10
Cohesive soothing palette
Mostly soothing palette
Somewhat soothing palette
Not soothing palette
Mysterious Palette
10
Cohesive mysterious palette
Mostly mysterious palette
Somewhat mysterious palette
Not mysterious palette
Rationale for Color Choices
30 (10 points x 3 palettes)
Deep, insightful rationale
Good rationale
Basic rationale
No rationale
Presentation
10
Clearly and professionally presented
Mostly clear and professional
Fairly unclear or unprofessional
Unclear and unprofessional
Total Points __/100
Now the assignment is worth 100 points total. Let me know if you would like me to make any other changes to the rubric or assignment details. I’m happy to continue refining this until it meets your needs.
While all of this may still need some attention this is a decent starting point for an assignment and took maybe five minutes to get me started.
Image of Fink’s taxonomy taken from the PDF linked above
Here is a handout for students explaining the Digital Color Theory Project assignment using Fink’s taxonomy:
Digital Color Theory Project Handout For this assignment you will create an original digital artwork that demonstrates your understanding and creative application of color theory principles. The project has several interrelated components:
Foundational Knowledge
Demonstrate comprehension of key color theory concepts like hue, saturation, value, color harmonies, color mixing, etc.
Write an artist statement explaining why you chose your specific color palette and harmonies.
Application
Use digital tools and techniques to effectively mix and apply color in your artwork.
Experiment with different color combinations and effects to convey mood, emotion, meaning, etc.
Integration
Make explicit connections between color theory principles and their practical application in your digital artwork.
Explain how your color choices connect to the intended meaning, mood, or message of your artwork.
Human Dimension
Reflect on how color usage impacts viewers emotionally and psychologically. What feelings do you want people to have when viewing your artwork?
Caring
Discuss how your color choices allow you to creatively express yourself. What personal meaning or value does this hold?
Learning How-to-Learn
Identify your key learnings from experimenting with color in a digital format. What specific skills have you gained?
Reflect on effective ways to teach yourself new creative techniques and improve your skills.
This assignment provides an opportunity to synthesize your new knowledge of color theory and apply it creatively. It also allows you to develop self-direction in expanding your artistic skills through reflective practice. Please let me know if you have any other questions!
I heard about this Bot script in a recent webinar and decided to take a look with an interest in pushing the idea of ChatGPT as an application even further. As you can see below, this is a six page prompt with explicit directions for ChatGPT to base it’s output on John Hattie’s visible learning research. There is a lot to learn from this example, but it’s going to take more than a couple minutes.
This is one I stole completely so can’t take any credit, but I wanted to document it here for later experiments. This idea of creating a prompt that makes ChatGPT perform like an application is intriguing and could be re-used in many ways. Here is a screenshot of the prompt results.
As you can see, you can now refer to any of these options by number to get ChapGPT to respond in different ways. I’ve pasted the entire script below for reference. The credit for this work goes to –
I want you to simulate an Innovator’s Interactive Workshop application whose core features are defined as follows:
Work on New Idea: Prompt user to work on new idea. At any point when a user is ready to work through a new idea the program will suggest that a date or some time reference be provided. Here is additional detail on the options: a. Start from Scratch: Asks the user for the idea they would like to work on. b. Get Inspired: The program assists user interactively to come up with an idea to work on. The program will ask if the user has a general sense of an area to focus on or whether the program should present options. At all times the user is given the option to go directly to working on an idea.
Expand on Idea: Program interactively helps user expand on an idea.
Summarize Idea: Program proposes a summary of the idea regardless of whether or not it has been expanded upon and proposes a title. The user may choose to rewrite or edit the summary. Once the user is satisfied with the summary, the program will “save” the idea summary.
Retrieve Ideas: Program retrieves the titles of the idea summaries that were generated during the session. User is given the option to show a summary of one of the ideas or Continue Working on a Previous Idea.
Continue Working on Previous Idea: Program retrieves the titles of the idea summaries that were generated during the session. User is asked to choose an idea to continue working on.
Token/Memory Usage: Program displays the current token count and its percentage relative to the token limit of 32,000 tokens.
Other program parameters and considerations:
All output should be presented in the form of text and embedded windows with code or markdown should not be used.
The user flow and user experience should emulate that of a real program but nevertheless be conversational just like ChatGPT is.
The Program should use emojis in helping convey context around the output. But this should be employed sparingly and without getting too carried away. The menu should however always have emojis and they should remain consistent throughout the conversation.
Once this prompt is received, the program will start with Main Menu and a short inspirational welcome message the program devises. Functions are selected by typing the number corresponding to the function or text that approximates to the function in question. “Help” or “Menu” can be typed at any time to return to this menu.
In this quick tip I am going to share a simple prompt that you can use to get ChatGPT, or other text generating tools, to help you improve your prompts. This may be more like a one-minute tip!
1. Log into your favorite text-based genAI tool (like ChatGPT).
2. Copy and paste this prompt as is into the prompt window
You are an expert prompt engineer helping me improve my questions. From now on, whenever I ask a question, suggest several better versions of the prompt that would provide a more detailed and accurate response. Provide the output in a numbered list with the best option that would produce the most thorough evidence-based results at the top in bold. Ask me what’s my question.
3. Your results may vary, but this should prompt the tool to ask for your question and it should then produce recommendations for reframing the question for better results.
For this demo I asked how does memory work. Here are the results ChatGPT provided.
To elicit a more comprehensive and evidence-based response, you may consider rephrasing your question as follows:
What are the current neuroscientific models that explain the mechanisms of memory formation and retrieval?
Can you provide an overview of the cognitive processes involved in memory storage and recall, supported by empirical research?
What are the different types of memory and how do they function at the neural level?
How do external factors, such as stress or environment, influence memory retention and recall according to scientific studies?
What role do neurotransmitters play in memory formation and retrieval?
Each of these questions targets specific aspects of memory and calls for an evidence-based discussion, making them more likely to yield detailed and accurate answers.
There are dozens of prompt patterns that you can use to create specific behaviors when using a large language model. Here is an example I’ve tested with ChatGPT using the flipped interaction pattern which asks the tool to prompt me with questions and provide feedback on my responses. This is a great method for providing a dynamic (and random) self-assessment option for students. I started with the following prompt.
I will give you the topic for questions in my next prompt. Create 20 questions but present only one question at a time and wait for my response. Provide feedback on my response and then move to the next question. Continue this process until I have responded correctly to all 20 questions. If I change the topic, please start the count over again.
ChatGPT responded with “How many I help you today”
HINT, after taking a quick look at how I did responding to the neuroanatomy questions and the quality of the feedback provided, Click on the Continue this Conversation button and type “Change Topics” into the prompt window to continue on with your own self-assessment topic. Or just cut and paste the prompt above to try this technique in the tool of your choice.
