The analogy of “ doctors suggest you get a second opinion!” is helpful. Thank you. I had not thought of using different AIs in that manner.
This is not a criticism; it’s an observation. After reading your dialogue with the AI, I get the feeling that using AI took longer and required more effort than if you’d simply sat down and written the post yourself. From reading the dialog script, I can see how it is useful for refining one’s thoughts or writing, but it also seems like it took a lot longer.
Also, are you familiar with the RTF Framework for AI prompts? From my understanding, rather than a long string of initial text for a prompt, as in your example, “best practice” is to break the prompt down into Role, Task, and Format. I read a book on AI recently that recommended this approach. Below is how it is described after doing a search.
- Role:
Define the persona or perspective the AI should adopt, such as “a financial advisor” or “a historian.” This helps set the context for the response.
- Task:
Specify the exact action the AI needs to perform, like “summarize this report,” “write a persuasive argument,” or “draft a formal letter.”
- Format:
Indicate how the output should be presented, including the style, tone, or structure, such as “a bulleted list,” “a conversational explanation,” or “a professional report.”
Example RTF Prompt:
• Prompt: “Take on the role of a history teacher. Task: Clearly explain the causes of the French Revolution. Format: Use a bulleted list highlighting the key points.”
This framework ensures clarity, alignment with your needs, and more refined AI-generated results.
This is how it seems to me as well, but reading through the text, I am wondering if there is some sort of physical challenge involved here? Something that would make it harder to sit down and write out a long post?
I have used this one for more academic research - I wish I could remember where I found it
PARTS
P, stands for persona. Persona refers to the person’s role. For example, I am a scholar.
A, stands for aim. For example, I want to write a paper on critical thinking.
R, stands for recipient. For example, it should be in a way that will convince SSCI journal editors.
T, stands for theme. Here you need to define the boundaries of the paper. You should specify which topics you want it to emphasize.
S stands for structure. You should specify what you want the output to be like. For example, how many pages you want it to be in bulleted form or paragraphs.
Re AI prompt frameworks: I don’t know if this truly captures ALL the prompt, but it does dive into a lot of them. Your Guide to Mastering All The Different Prompt Frameworks (From the Juuzt AI website.)
A lot of them are really just variations on a theme, so perhaps we need a framework of frameworks. :wink
PS: I’m tempted by The Chain of Destiny Framework just for the name.
Ha! Yes! I know … it does look like that!
But … I have no doubt that it was waaay quicker for me to work this way than sit and type it out*.
I used to be an “estimation expert”! And one lesson I learnt from that is that we humans are terrible at estimating how long it will take to do something. Another was that we underestimate, especially when someone else is doing the work.
My thoughts - off the top of my head, and typed in about 45 minutes (like a Luddite) follow:
Effort / Perception:
-
Whereas you read my first long garbled request, I actually spoke that out loud for approximately 10 minutes. ChatGPT cleaned my garbled words up in about 20 seconds. That “cleaning” probably saved me an hour. That time saving won’t be obvious, but it was massive for me (and for anyone, I think).
-
When you were reading the script back … you had to read / scan through are an awful lot of boring repeated words. ChatGPT repeats back loads and loads of words in his responses. I didn’t have to reread most of them - I scanned for the changes.
The effort to read the dialog back might skew the perception of the effort required to “write” it.
Comparison
- Some people are very fast writers compared to others.
- I’m not a fast writer, but then have published 5 books that have done okay, so I must be okay at.
- It’s taken me about 45 minutes to type and edit this. But, of course, I wrote a lot more words than you’re seeing here, but they’re long gone!
Hope that helps. There’s no black and white answer, but it was definitely easier. That said, it also think it was definitely easier to sit and “think with my finger tips” for this response, though I could be completely wrong because I didn’t try the other way.
- Especially considering that I wouldn’t have written it if I’d had to type it. Before I sat down to write, I weighed up the effort involved in sharing my experiences to help others, with the risk of being politely “beaten up” by people for openly using AI, and I decided it was worth it.
WRT to the prompts. I have told chatGPT and Claude who I am and how to work with me in their special instructions that kick off every chat. They include stuff about my ambitions, who I am, what I’m good at and not good at, my writing style, and stuff like that.
So it’s “prompted” behind the scenes. A bit different to prompts where you ask chatGPT to pretend it’s the love child of Gladwell and Hemingway … because I’m asking it to write and think like me.
I can share them, if you like. I don’t think there’s anything in there that’s secret!
You need not feel that way. This forum is one of the few places where you can freely share without fear of being flamed, at least most of the time. Keep writing and sharing. Only those who are “Proud Peters” are unwilling to learn and consider the merits of the opinions and practices of others.
“Proud Peter” comes from this excerpt of the book, * Humilitas*:
Perhaps the most obvious outcome of being humble is that you will learn, grow and thrive in a way the proud have no hope of doing. The logic is simple: people who imagine that they know most of what is important to know are hermetically sealed from learning new things and receiving constructive criticism.
I see this at conferences all the time, whether in business, education or not-for-profit settings. Every conference seems to have a Proud Peter. He’s the guy in your organization who is moderately talented and charming but whose years in the business have created an inflexibility when it comes to learning from others or implementing changes. His natural wit is able to point out the smallest difficulty with a new idea, and so he quickly convinces himself and sometimes others that the old way—his way—is probably best.
Dickson, John. Humilitas: A Lost Key to Life, Love, and Leadership (pp. 116-117). Zondervan. Kindle Edition.
I will echo @brookter 's comment. I think disclosure will be an increasingly important tool for building trust.
We’re moving our website to a new platform (WordPress → Astro). Along the way, we will add an AI disclosure post.
As the rate of LLM-generated slop increases, I think there will be real value in distinguishing yourself from the masses. (originally mistyped as the mEsses
Love that!
And the gist of adding a disclosure stuff. I guess- for me - it’s a chance to emphasise how using AI helps me help loads more people. There’s a taint to AI at the moment that comes from so much of the content we see being written by people who want to sell or influence or manipulate. In the past adverts used to be far more obvious.
About 5 years ago I’d go for a ride on my bike and get passed by all these people with white hair. Made me feel slow and bad! Then I realised they were riding ebikes, and I disliked them for “cheating”. Then I spoke to one of them and he said that the ebike had opened up his life enormously. He could ride further, and feel safe. He could go out on Windy days. He lived on a hill, so he was able to go out for a lovely bike ride almost every day, where as before he used to have to load his bike onto his car, drive somewhere and then … you get it.
Once I heard all that, Mark, I stopped think he was cheating at cycle. I thought he was outwitting the weather and age. And I bought an ebike and I go out on it almost every day, and it’s been a genuine life changer.
Like wise … just as some people will (I guess!) use ebikes and scooters to rob people and make fast getaways … most people won’t, they will use them for good.
The bias is really good way of describing it … in my biased opinion! Your biases might have different opinions
I can’t resolve this dilemma, I’m happy with my choice, for now.
Hey, if anyone is reading this and would like to read the book for free, email me at clarke@clarkeching.com and I’ll ping you a pdf version. It’s still being proofread, so it will be ever so slightly different to the version that gets released on the 12th.
You can take a look at it and judge whether it feels like it’s a good book, or not, despite me paying a proofreader to proofread it and not doing all that by myself.
BTW: Out fo curiosity I passed the book through grammarly’s AI checker and it spotted 3 paragraphs that it said might have been written by AI. I actually wrote those 3 100% all by myself. So, it’s possible that I’m a robot.
The value we ascribe to things includes the feelings inferred from context around the thing, like the story of their creation/creator…
This can lead to interesting distortions, like people or even brands constructing false, misleading or incomplete origin stories for their works, or themselves. (e.g. many influencers)
Now with the advent of generative AI, we live in very interesting times. If we were all swamped with content before, original creators will soon be overrun by boundless hoards of zombie content. As soon as someone works hard to create something new, the barriers to plagiarism have dropped to zero.
That was soooo good! Thank you.
I’ve just reworked my amazon blurb after reading that (or, I should say, with delight, that Claude.ai and I reworked it).
It now starts like this:
The Bottleneck Detective: Fix Bottlenecks, Boost Productivity, and Unlock Profits—in Just One Hour
From Clarke Ching, bestselling author of “Rolling Rocks Downhill” and “The Bottleneck Rules,” comes a revolutionary guide that distills three decades of Theory of Constraints expertise into one powerful hour of reading.
What if one hour could completely change how you solve business problems?
The bottleneck concept revolutionized manufacturing in the 1980s through Eli Goldratt’s The Goal. But today’s bottlenecks don’t just lurk in factories—they’re everywhere, from coffee shops to hospitals to remote work teams. After spending 30 years helping organizations worldwide unlock their potential through Theory of Constraints, I’ve crafted The Bottleneck Detective to bring these powerful ideas into the modern workplace in their purest, most accessible form.
The ted talk helped me realise that I hadn’t written anything into the blurb (which is my sales page) that calls out credibility and the huge amount of effort that’s gone into this.
(Funny thing: Last year someone stole Bottleneck Rules book and published it with a new title and a (surprisingly poor) sales blurb on amazon. So that book has been faked, just like that painting was!)
Thanks again Robert! Such a good video.
For anyone who’s been following this thread, you might like to know that I’ve just got my final document back from my proofreader.
The first thing I noticed, was that she had made far, far fewer changes (corrections) than with previous books. And … any typos came from me when I hand-edited. So that’s a huge bonus for me. And, for her, she charges by the word count, not by the time it takes her to do her work, so we are both happy.
The second thing - and I think this is important - was her comment:
Your writing style is truly extremely intriguing and refreshing. I attempted not to make any formidable changes to your style, voice, and tone but to retain your unique and compelling writing style, which is an absolute winner (especially in today’s AI era where every book sounds the same!).
I shared this because I know that most people think that AI is going to produce bland writing. It doesn’t have to be that way.
On a Mac specific topic: I’m now using Vellum.app to prepare the kindle and paperback files. It’s a magnificently powerful and easy tool, that’s such good value for someone like me. It’s saved me hours and prevented a load of stress.
Hope this stuff helps.
I like how you manage so many subscriptions, very inspiring.
So I’m thinking of keeping PerplexityAI because:
- my content creation (non-fiction articles, blogs, books, speeches and trainings) always start with an exploratory phase, and PerplexityAI is very good at it
- for small tasks I can get the same result as the original models (GPT, Claude…)
In the meanwhile I’ll circle around Claude and ChatGPT only when I really need to do the hard writing (writing books is not my main job, I write books and articles mainly to spread information about my topics).
Inspired by you, probably there will be periods when I’ll keep only the subscription of PerplexityAI.
I like how you manage so many subscriptions, very inspiring.
It’s funny how reluctant I was to pay for Claude when I already had ChatGPT, but now - while I’m energetically writing - they’re an absolute bargain. And I can turn one of them off whenever I want to.
Just now, I’m focused on making money, not saving it.
I also have perplexity, but I somehow wangled a free year on pro. I like it a lot.
My book published on Amazon over the weekend and I’ve started getting a few notes from strangers who’ve read it.
Some of the comments are about the content, but a few have been about the writing.
As you can imagine, I’ve been feeling nervous about how people will react. The content and ideas, and the choice to write with a detective voice, were all obviously entirely mine, but I put a load of effort into woking with ChatGPT and Claude to create the “Chief Detecive Clarke Ching” voice. I didn’t want to embarrass myself, and I want to help people, and I’m not especially thick skinned, so … well, you get it.
Anyhow, here’s the first note from a stranger, and he said, “I love the writing style”. I think the extra effort paid off:
I know AI is already helping a lot of / content farmers produce even more , even faster.
But, I wanted to share my example, because I’ve spent 25 years teaching myself to be a good writer, and this helping me more good stuff, faster.