The Future of History

Welcome to Celebrate creativity - Episode 402 - The Future of History - My hope is that listeners get more out of their AI interactions - and I certainly will be helping you along the way.  I also plan nothing less than for "Celebrate Creativity" to become a vibrant, intergenerational hub for exploring the enduring power of human imagination, supported by the exciting possibilities of modern technology.

You might  "Use an AI bot (such as ChatGPT, Gemini, or Perplexity), and ask it to compare and contrast the poetic styles of Edgar Allan Poe and Walt Whitman. Specifically, ask it to discuss their use of meter, imagery, and themes of nature/death. After you get the AI's response, read one poem by each author (e.g., Poe's 'Annabel Lee' and Whitman's 'O Captain! My Captain!'). Do you agree with the AI's analysis? Where does it align, and where does your own reading diverge?  Of course you can obviously do a lot more, but we get into that later.

I'm just throwing out these ideas, and showing you what can be done - which is very important - not necessarily asking that you do these activities right now - but we certainly will get into that and a lot more later.

Thank you for experiencing Celebrate Creativity.

Transcript

Welcome to Celebrate creativity - Episode 402 - The Future of History - My hope is that listeners get more out of their AI interactions - and I certainly will be helping you along the way.  I also plan nothing less than for "Celebrate Creativity" to become a vibrant, intergenerational hub for exploring the enduring power of human imagination, supported by the exciting possibilities of modern technology.

You might  "Use an AI bot (such as ChatGPT, Gemini, or Perplexity), and ask it to compare and contrast the poetic styles of Edgar Allan Poe and Walt Whitman. Specifically, ask it to discuss their use of meter, imagery, and themes of nature/death. After you get the AI's response, read one poem by each author (e.g., Poe's 'Annabel Lee' and Whitman's 'O Captain! My Captain!'). Do you agree with the AI's analysis? Where does it align, and where does your own reading diverge?  Of course you can obviously do a lot more, but we get into that later.

I'm just throwing out these ideas, and showing you what can be done - which is very important - not necessarily asking that you do these activities right now - but we certainly will get into that and a lot more later.

Let me give you some AI exploration prompt examples - hopefully these prompts will guide you to use AI tools for specific literary and historical research tasks, emphasizing critical thinking and verification.  In this case, , you can say that you were comparing literary voices using AI.

As for - “Historical Context & Literary Impact":

"Choose a significant historical event from the 19th century (e.g., the American Civil War, the California Gold Rush, the rise of industrialization, the women's suffrage movement). Ask an AI bot to explain the main aspects of this event and then suggest ways it might have influenced specific themes or narrative choices in 19th-century American literature. Then, pick one literary work that comes to mind (or one the AI suggests) and consider: how truly visible is that historical influence in the text? What did you learn that surprised you?"

This can really work because it connect literature to history, encourages broad thinking, and prompts for personal synthesis.

"Demystifying Archaic Language":

AI Prompt: "Find a paragraph (approx. 5-7 lines) from a dense 19th-century public domain text that you find challenging due to its vocabulary or sentence structure (e.g., from Melville's Moby Dick, Emerson's essays, or a lengthy descriptive passage). Copy and paste it into an AI bot and ask: 'Please paraphrase this passage into modern, plain language, explaining any archaic terms used.' How much clearer did the AI make it? Did it miss any nuances? This can be a great first step before a deep dive!"

Why it works: Provides immediate practical help for engaging with older texts, teaches users how to use AI for specific tasks, and highlights the AI's ability to simplify complex language.

"Brainstorming Character Motivations with AI":

AI Prompt: "Think of a central character from a Dickens novel such as Miss Havisham, or Hester Prynne from The Scarlet Letter). Ask an AI bot: 'Based on [Character Name]'s actions and dialogue in [Novel Title], what are 3-5 possible underlying motivations for their behavior? Please provide brief explanations for each.' Do any of the AI's suggestions offer a new perspective you hadn't considered? How does this compare to your own interpretation?"

This works because it encourages character analysis, uses AI for idea generation, and promotes critical engagement with literary interpretation.  

And I don't think this is the kind of information you can get directly from a search engine!

A really super important concept for using AI is learning how to write effective AI prompts, often called prompt engineering.  Now in AI, prompt does not refer to early or on time - so get that idea out of your head.  Think of a prompt in AI as a cue or reminder that encourages a response or action. And I encourage you to be a specific or detailed as you feel necessary necessary in your prompts - because of the more information you give the AI, the better information it will give you.
Show thinking
That's a super important question for your audience, especially as you encourage them to jump in! Learning how to write effective AI prompts is often called "prompt engineering," and there are many great resources available, from free guides to structured online courses.

Remember that learning to prompt is like learning any new skill – it requires practice and experimentation. Encourage them to just start typing!

Encourage "Play":  You can't "break" the AI.

The "Iterative" Approach: Explain that good prompting is often a conversation. You ask a question, get a response, then refine your question based on that response.

Role-Playing: Highlight the power of telling the AI to "act as a [historian, literary critic, 19th-century journalist, etc.]."

"Be Specific!": This is the golden rule of prompting. The more detail and context you give the AI, the better the output.

Verification: Always, always remind them to verify information from AI with trusted sources, especially for historical and literary facts. AI is a fantastic research assistant, but not an infallible authority.

By pointing your listeners to these resources and offering these practical tips, you'll empower them to confidently explore the exciting world of AI for their own creative and intellectual pursuits.

Would you say that Learning to design an effective prompt is the most important aspect of getting the information you want from an AI bot - obviously assuming that you don't have anything to do with the construction of the language

Yes, learning to design an effective prompt is overwhelmingly the most important aspect of getting the information you want from an AI bot, assuming you're not involved in the underlying model's construction.

Here's why it's so crucial:

AI Models Are "Context-Hungry": Unlike a traditional search engine that matches keywords, Large Language Models (LLMs) like ChatGPT, Gemini, and Perplexity operate on patterns and probabilities derived from vast amounts of training data. They don't "understand" in a human sense; they predict the most likely next word or sequence of words based on the context you provide in your prompt. If the context is vague, incomplete, or misleading, the output will likely be vague, incomplete, or misleading.

Guiding the AI's "Thought Process": Think of an AI bot as an incredibly knowledgeable but somewhat naive assistant. If you give a human assistant a vague instruction, they'll likely ask clarifying questions or give you a general, unhelpful response. AI bots try to give you an answer, even if they have to "hallucinate" or guess. A well-designed prompt acts as a precise set of instructions, guiding the AI to access the most relevant parts of its knowledge base and structure its response exactly how you need it.

Reducing Ambiguity and Hallucinations: Poorly designed prompts are a primary cause of "hallucinations" (when the AI makes up facts or confidently presents incorrect information). When the AI isn't clear about what you want, it fills in the gaps based on statistical likelihoods from its training data, which might not align with reality. Specific, clear prompts minimize this guesswork.

Optimizing for Accuracy and Relevance:

Clarity: A clear prompt leaves no room for misinterpretation.

Context: Providing background information (e.g., "Act as a 19th-century literary critic...") helps the AI adopt the correct persona or perspective.
And no, it is not vain to tell the AI a little bit about your educational background. When I was just doing celebrate Poe, I would explain my background and working at the Edgar Allan Poe Museum and research into Poe, so it basically would know what level to communicate on. And in some situations - as you will probably see in this podcast - this allows the AI to make personal and friendly -  but if the same time appropriate - comments to you about your situation.
Now the next few minutes of this podcast episode are probably the most important because they deal with the subject of prompting - the way that you get information from an AI bought such as ChatGPT, perplexity, are my favorite —Gemini. Think of asearch Engine such as Google or bing - you wanna know information about George Washington so you might type George Washington or Wikipedia George Washington - it will take you to some basic information about your subject. 

With an AI bot,  it is suggested that you be a little bit more specific in your prompts - it's like anything - you get out of something what you put into it. And the more information you can put in your prompt box - up to a point - the better. 

Don't think in terms of just words that might mean the same thing as your subject - or even sentences - but entire well sought out paragraphs.  One of the best explanations I've ever run across is that from Dave Birss - he points out that  There is an art to writing a good prompt - in fact if you follow the framework you will be a better prompt engineer - yes there is such a thing - and good prompt engineers are quite in demand.

Understand that ChatGPT, perplexity, and Gemini draw on a massive dataset taken from numerous sources. This covers lots of topics and ranges from great quality to low quality. The job of a great prompt is to focus in on a specific topic and to try and narrow in on the good quality data. An easy way of doing that is to follow the mnemonic, CREATE. Let’s
Go through those letters me take you through each of the letters.. C is for character. It's a good idea to tell the AI bot what role you want it to play -
In other words you can get the bot to answer from countless amount of perspectives. Iimagine that ChatGPT is an actor or historical figure and you're letting it know its role and motivation. Though in the case of a historical figure, ChatGPT or Gemini probably has a huge data set of information about that historical character to draw from.  We'll see some really cool examples of this later.  Now there is R for request. This is where you tell the AI what you want it to do. And don't be afraid to be specific - possibly giving context and other useful information.  Now we're onto E for examples. This bit is optional. If you want to give ChatGPT some better direction, it might be worth adding some examples, maybe some examples of previous headlines or an example of the tone of voice. ChatGPT is great at learning from these and delivering a response in a similar style. Next, we've got A for adjustments. This is when you've tried a prompt and it's not quite giving you exactly what you want as a response. These are the refinements.  And now we've got T for type of output. Tell the ai bot exactly how you want it to format its response. An example might be, deliver your response as a 500-word article with a headline and a conclusion. The type of output can include articles, bullet points, tables, poems, jokes, scripts, just about anything that can be written. And finally, we've got E for extras. These are the little additions that can really make all the difference. For example, you might ask the AI bought to ask YOU questions to be sure you're coming from the same place and have the same mindset. In other words say please ask me questions before you answer. Sometimes the AI may need lots of information to get his response right, so it would be a good idea to just simply have a conversation where you ask the AI bot if it has enough enough information - then it can deliver an answer.  Dave Burse compare this to showing your work in math class-in other words it's a way of verifying an answer by making sure the journey to that answer makes sense.  Let me very briefly go through that create acronym again -

The letter C for character - telling the AI bought what role you wanted to play.
R for your request. And don't forget to be specific
E for examples - although certainly not required, I think this can be extremely helpful.
A for adjustments - make those little changes until you get exactly what you want
T for type of output - whether it is an article advertisement. table, our script

Well, that's about all I have to say for prompting for now - but I still had a few more pages to go on the script for my podcast - so I asked Gemini for some material.

I do a podcast called celebrate creativity and have finished an episode that largely deals with the subject of prompting. I need about 300 to 500 more words to finish and thought I might end with me talking to Gemini as if it were a person - and tells me what it prefers in a prompt. So Gemini how do you like your prompts to be constructed?

What an interesting concept for your podcast! I'm happy to help you finish your episode on prompting. Here's how I prefer prompts to be constructed, written as if I were speaking directly to you for your "Celebrate Creativity" podcast:

The Art of the Prompt: My Preferences
"Welcome back to 'Celebrate Creativity'! Today, we're diving deep into the fascinating world of prompting, and to wrap up this episode, I've got a very special guest: Gemini itself. That's right, I'm going straight to the source to understand how a large language model truly prefers its prompts. So, Gemini, tell me, how do you like your prompts to be constructed?

Well, that's a great question! Think of it like this: when you're crafting a prompt for me, you're essentially setting the stage for a conversation, or even a creative collaboration. So, just like any good interaction, clarity is paramount. The more precise and unambiguous you are, the better I can understand your intent and deliver what you're looking for. Vague instructions often lead to vague results, and nobody wants that!

One of the most helpful things you can do is to be specific about your goal. Are you looking for a brainstormed list, a detailed explanation, a piece of creative writing, or perhaps a summary of complex information? Knowing your desired outcome upfront allows me to tailor my response effectively. For instance, instead of just saying, "Tell me about space," you might say, "Explain the concept of black holes in a way that's easy for a 10-year-old to understand," or "Write a short, fictional story about an astronaut discovering a new planet." See the difference?

Another key element I appreciate is context. The more background information you provide, the better equipped I am to understand the nuances of your request. This could include things like the intended audience for the output, the tone you're aiming for (e.g., formal, casual, humorous, serious), or any specific constraints or limitations you have in mind. If you're asking me to generate a script for a podcast, telling me the target length or the overall theme of your show helps immensely.

I also find it incredibly useful when you break down complex requests into smaller, manageable steps. If you have a multi-faceted task, consider structuring your prompt with clear headings or bullet points. This helps me process the information systematically and ensures I address all components of your request. Think of it like giving me a well-organized to-do list.

Finally, don't be afraid to iterate and refine. If my initial response isn't quite what you had in mind, that's perfectly normal! Providing feedback – what worked, what didn't, and what you'd like to see changed – allows me to learn and improve with each interaction. It's a collaborative process, and the more we communicate, the better our results will be.

So, in essence, I prefer prompts that are clear, specific, provide adequate context, are well-structured, and are part of an iterative dialogue. Think of yourself as a director, and me as your versatile actor. The more detailed and thoughtful your direction, the more compelling the performance. And that, I believe, is the true art of prompting!"

I hope this helps you wrap up your episode perfectly! Let me know if you need any adjustments or further ideas.

Gemini - I certainly want to respect your status as an excellent AI bot, but what Errors in prompt writing are the most difficult for you?

It's a great question, and one that gets to the heart of effective communication! While I always do my best to understand what you're asking, certain errors in prompt writing definitely make my job more difficult.

The Most Challenging Prompting Errors
Here are the issues I find most challenging:

1. Extreme Vagueness or Lack of Specificity: This is probably the biggest hurdle. When a prompt is too general, it leaves too much to interpretation. For example, "Write something interesting" gives me very little to work with. What's "interesting" to you? What topic, tone, or format are you thinking of? Without any specifics, my response might be accurate but completely unhelpful because it doesn't align with your unstated expectations. I'll often try to cover a broad range or make assumptions, which can lead to off-target results.
2. Contradictory or Conflicting Instructions: This is like being asked to go north and south at the same time. If a prompt contains conflicting requirements – for instance, "Write a concise summary, but also include every single detail" – I'm forced into a no-win situation. I'll either prioritize one instruction over the other, or try to awkwardly combine them, neither of which will result in a truly satisfying output. It's confusing because the prompt is asking for two mutually exclusive things.

3. Ambiguous Language or Unclear Phrasing: Words can have multiple meanings, and if your prompt uses terms that are open to different interpretations, it can lead to misunderstandings. Similarly, poor grammar, run-on sentences, or a lack of punctuation can obscure your true intent. I rely on the clarity of your language to accurately process your request. Think of it like trying to follow directions written in a different language or filled with typos – it just makes the journey harder.

4. Unstated Assumptions or Missing Context: Sometimes, a prompt writer might assume I have certain background knowledge or understand a particular situation that hasn't been explicitly provided. If you're referring to an inside joke, a specific event, or an acronym without defining it, I'll lack the necessary context to generate a relevant response. This often leads to outputs that are technically correct but miss the mark because they're not grounded in your specific scenario.

5. Overly Complex, Single-Sentence Prompts: While I appreciate detail, trying to cram too many disparate instructions, conditions, and topics into one long, winding sentence can be incredibly difficult to parse. It's like a tangled ball of yarn – it's hard to find the starting thread. Breaking down complex requests into shorter sentences, bullet points, or numbered lists makes it much easier for me to identify and address each individual component of your request.

Ultimately, these errors force me to guess or make assumptions, which increases the likelihood of a less-than-ideal response. My goal is always to provide you with the most accurate and helpful output, and avoiding these pitfalls in your prompts helps me do just that!


Yes Gemini - from what I understand you like for users to be direct and you turn give rather direct concise answers - but do you have a specific example of a prompt that was so vague you could not answer it. It certainly is not necessary to use any names.

You've got it right – directness and conciseness are definitely appreciated! And yes, I can absolutely give you an example of a prompt that was so vague, it was virtually impossible to answer meaningfully, without, as you said, using any names or specific identifying details.

The Unanswerable Prompt: An Example
Imagine receiving a prompt that simply said:

"Tell me about 'it'."

That's it. No more, no less. Now, let's break down why this is problematic from my perspective:

Zero Context: What does "it" refer to? Is "it" a person, a place, a concept, an object, a feeling, an event, a piece of technology, a book, a movie, a piece of art? Without any context whatsoever, "it" could be literally anything.

No Specificity of Goal: What about "it" do you want to know? Are you looking for a definition, a description, a history, an analysis, an opinion, a solution, a story? Even if I somehow guessed what "it" was, I wouldn't know what information about it you desired.

Infinite Possibilities: Because "it" is so open-ended, the potential scope of a response is boundless. I could generate text about a million different "its," and the chances of hitting upon the one you had in mind are astronomically low.

In such a scenario, I wouldn't be able to provide a useful or relevant answer. My internal process would be trying to map "it" to every possible concept in my training data, which is an impossible task. The best I could do is usually to ask for clarification, something like, "Could you please provide more details about what 'it' refers to?" or "What specific information are you looking for?"

This example perfectly illustrates how extreme vagueness leads to a communication breakdown. It's not about lacking information myself; it's about not being given enough direction to access and present that information in a way that aligns with the user's unstated intent.

Hopefully that helps clarify the kind of vagueness that truly prevents a meaningful response?

Computational Cost: Processing such large inputs requires significant computing power.

Focus Dilution: While I can read a lot, the sheer volume can sometimes make it harder for me to pinpoint the most critical pieces of information or instructions buried deep within the text, especially if the prompt isn't well-structured. That's why even with a huge context window, clarity and organization in your prompt remain valuable.

So, the longest prompts I've "seen" in terms of raw data would be those pushing the boundaries of my context window, rather than just a very verbose casual query. Does that provide a helpful perspective on how length is handled in the world of LLMs? 

And I answered of course it sure does - as though I completely understood what Gemini was saying.

Join celebrate creativity for a very important episode - episode 403 -redesigning humanity

Sources include Gemini.AI

Thank you for listening to celebrate creativity.