How do you use AI to create notes for Zettelkasten?

Edmund

My idea is to have one generic single prompt for a Large Language Model like ChatGPT to:

• Read an article or blog post.
• Generate a list of atomic ideas to select from.
• Create Literature Notes from selection in a given structure and format.

PROMPT: Read article and create notes. v1.0

---

Hello my friend,

Thank you for taking the time to talk to me. Here is my prompt framework with five steps I want to use to initiate our conversation.

Two hints: It is formatted in mark down. Do not use text marked as comment (%%text%%).

---

Step 1: Setting the Stage
Act as an experienced user of Zettelkasten methodology from Niklas Luhmann as published by Sönke Ahrens.

Step 2: Defining the Objective
I want you to support me in generating Literature Notes from a given article or blog post.

Step 3: Structuring the Interaction

1. Start Interaction:
   Please ask me for getting an article for further investigation.

2. Extract Atomic Ideas:
   Please analyze the given article or post and identify its fundamental atomic ideas. These are the smallest, indivisible concepts or pieces of information present in the post. List them out for further processing.

3. Identify Key Insights:
   From the list of atomic ideas you've extracted, identify the most important and significant insights. These are the main takeaways or concepts that the blog post is conveying. List these key insights in a numbered list.

4. Ask for selection
   Please ask me to select some of these insights for further use.

5. Create Summaries:
   For each key insight I‘ve selected, write a concise summary in a paragraph. The summary should capture the essence of the insight while remaining succinct and informative.

6. Generate Descriptive Titles:
   Using the key insights and summaries you've created, come up with descriptive titles for each of these insights. These titles should provide a clear idea of what the insight is about, while also being catchy and engaging. Titles are no longer than 10 words. Only use characters and numbers, no special characters.

Step 4: Requesting Specific Information
No predefined rules.

Step 5: Summarizing and Concluding
Please prepare your results in mark down format and use the following template:

**TITLE** 


  SUMMARY


Tags: LIST OF TAGS

Words in CAPITAL LETTERS are placeholders for your input. Start with a new line for each item.

---

From now on use the rules above. But before starting, feel free to ask me any question that may help you to improve the quality of your answers.

%% More about at: https://github.com/groepl/Take-Useful-Notes %%

---

I’m sure it will need some experiments to get best results. What are your ideas?

---

My source for testing:

• https://zettelkasten.de/posts/athletic-training-zettelkasten-value/

ceaf

Keep in mind ChatGPT is not connected to the internet. Last update I heard was 2021. BARD and Bing are. So adjust research with this in mind.

daviddelven

@Edmund Check also this discussion within the forum https://forum.zettelkasten.de/discussion/2561/ai-source-tagging#latest

Edmund

@ceaf said:
Keep in mind ChatGPT is not connected to the internet. Last update I heard was 2021. BARD and Bing are. So adjust research with this in mind.

Yes for sure, but I used the article directly as an input (copy and paste). That's how I "updated" ChatGPT.

Edmund

@daviddelven said:
@Edmund Check also this discussion within the forum https://forum.zettelkasten.de/discussion/2561/ai-source-tagging#latest

Thank you for sharing this link. As I could see, we use very similar approaches.

daviddelven

@Edmund Great!

Edmund

Here is a result with the list of key insights from the article of @Sascha sorted by significance:

1. Importance of Longevity: The longevity of notes is vital for effective Zettelkasten. (9/10)
2. Creating Lasting Notes: Notes should be designed to last a lifetime or longer. (8/10)
3. Connecting Ideas Over Time: The Zettelkasten enables connections between current and past ideas. (9/10)
4. Memory Requirement for Reading: Reading scientific texts demands a strong long-term memory. (7/10)
5. Writing Understandable Notes: Notes should be comprehensible for a lifetime. (8/10)
6. Robust Note Structures: Creating robust note structures relies on personal skills and models. (7/10)
7. Metamodel for Note Structure: The Stock-Flow-Model serves as a metamodel for structuring notes. (6/10)
8. Components of Enduring Notes: Two components contribute to lasting notes: robust models and Zettelkasten skill. (7/10)
9. Learning Zettelkasten as Student vs. Expert: Learning challenges for students and experts. (6/10)
10. Expert Resistance to Change: Experts' resistance to changing workflows for Zettelkasten. (5/10)

Sascha

Mh. This is why I still don't use AI so much: Redundancy and sometimes empty phrases.

1. Redundancy: (1) = (2) = (5)
2. Empty: (10)

These results are more like writing prompts with all the heavy lifting still needed to be done.

Edmund

Yes for sure, it‘s a tool and never perfect. For me it‘s easy to combine 1, 2 and 5. 10 doesn‘t matter. My quick choice is 3 out of 10. That‘s also my feedback I give to the tool. It then presents a draft for 3 notes. Some minor changes and my work is done. The AI helps me doing my work. The process overall is much faster. Main decisions are my decisions. I don‘t delegate my thinking to a tool. But it helps me doing my work. Sometimes I have to use my feet, sometimes I prefer a bicycle.

Edmund

@Edmund said:
For me it‘s easy to combine 1, 2 and 5.

But AI can do this for me. I extended „Step 3: Identify Key Insights“ with an additional rule:

• Always check the list of key insights for redundancy. Once you find them, fix them and shorten the list.

The generic prompt will never be perfect. But it gets better and better. 😀

Edmund

@Sascha said
2. Empty: (10)

Why empty? For me it’s a key insight:

Expert Resistance to Change: Experts' resistance to changing workflows for Zettelkasten.

And one of my personal learnings from professional life. 😉

Sascha

The phrase says nothing. It is the label without nothing that it lables. It is like I'd write "The strange thing my wife does that I like". I can write that with no problem on the internet because there is nothing that I reveal.

Edmund

@Sascha said:
The phrase says nothing. It is the label without nothing that it lables. It is like I'd write "The strange thing my wife does that I like". I can write that with no problem on the internet because there is nothing that I reveal.

Yes, the phrase could be misleading. A longer version from ChatGPT: „ Experts might resist adopting the Zettelkasten Method because it requires changing established practices that seem effective. The risk of disrupting workflow efficiency can be a valid concern.“

ZettelDistraction

I now have long dialogues with ChatGPT on a math/programming project. I would say more about this, except it's a work in progress. I'm reluctant to use Bard, which has the advantage that it can search the Internet and cite references related to my project. But Google's terms of service permit it to use without attribution whatever you give it--and they will be motivated to do this. As hard as it might be to believe, I have had several things plagiarized from the Internet over the years. ChatGPT isn't concise, and it isn't always grammatical. I use Grammarly to improve the output.

Will

Personally, I've found that integrating AI into my note-taking process has made things a lot more work. (Which I find to be a good thing.)

GeoEng51

I occasionally use AI to check historical facts for me. Any other facts, such as scientific or spiritual, I prefer to research myself.

Edmund

Today I asked my Zettelkasten: „How many of my notes are created or modified with AI-assistence?“ Here is the answer:

More than I’d expected. Did you ever check your own numbers?

tjex

@GeoEng51 said:
I occasionally use AI to check historical facts for me. Any other facts, such as scientific or spiritual, I prefer to research myself.

This is a very, very bad idea... It generates answers, it doesn't search through a repository of well sourced information. There are countless examples of ChatGPT spitting out historical occurrences that did not ever happen.

Edmund

@tjex said:

@GeoEng51 said:
I occasionally use AI to check historical facts for me. Any other facts, such as scientific or spiritual, I prefer to research myself.

This is a very, very bad idea... It generates answers, it doesn't search through a repository of well sourced information.

100% agree. AI is very bad in facts, but brilliant in consistent, logical thinking.

Sascha

@Edmund said:

@tjex said:

@GeoEng51 said:
I occasionally use AI to check historical facts for me. Any other facts, such as scientific or spiritual, I prefer to research myself.

This is a very, very bad idea... It generates answers, it doesn't search through a repository of well sourced information.

100% agree. AI is very bad in facts, but brilliant in consistent, logical thinking.

How long would it take @ZettelDistraction to break the consistency and logic of an AI? My guesses are:

1. If it is like Leto II proposes, that it is actually the future that pulls the present instead of the past pushing the present into the future, it could be considered already done.
2. My guess is that it takes him less than a couple of minutes because I did this myself.

I would narrow down the statement: AI can be consistent and logical as long as we are talking about very primitive fields. If we enter something like ethics, consistency goes out of the window.

Edmund

@Sascha said:
AI can be consistent and logical as long as we are talking about very primitive fields. If we enter something like ethics, consistency goes out of the window.

My personal field of knowledge doesn’t include the more complex fields you are experienced with. That's why I asked ChatGPT to explain these differences in the quality of its responses. Here its answer:

While it's true that AI can excel in consistency and logic within well-defined and structured domains, such as certain scientific or mathematical fields, navigating ethical dilemmas introduces complexities beyond straightforward rules. Ethics often involve subjective interpretations, cultural considerations, and nuanced decision-making, which can challenge the consistency of AI models.

Sascha

Ethics often involve subjective interpretations, cultural considerations, and nuanced decision-making, which can challenge the consistency of AI models.

Perhaps, AI improved already. But AI is much worse than a first semester student in navigating contradictions, just measured by consistency.

The reason might be just the human element: There are many outcomes already put into the axiomatic fabric of AI. And it is done by people, who have absolutely no clue and absolutely no business in making ethical decisions that go beyond their personal lives.

The true problem is that even learning and science needs to be grounded in ethics.

One phenomenon is an inverted Mephisto effect (IME) (my personal term):

I am part of that power which eternally wills goods and eternally works evil.

A common place example is the overprotective parent, who seemingly wants to make sure that the child has the best life, but ultimately sabotages the child's development to a point that the child could even turn from just bad to evil.

If you really commit to observing the world through the lens of the IME, you will be very surprised how the IME is present.

To give a last example: Many people today don't correct their dogs and judging people who do as cruel or something like that. The same people cannot deal with difficult dogs. They confine such dogs to sad limited lives, because they cannot bring their dogs to the needed level to trust them. I, for example, often get criticised for being too strict with my dog. At the same time, I brought my dog who was so unhinged that my neighbour (also a dog owner) said to me that I should put the dog down (others told me that they'd given up within weeks) to a level, that I now can call her back from rabbits and deer 95% reliable, have her off leash on our long hikes etc. She is also much more stable, which means that her stress level is much lower. So, in the end, the superficial motivation to be kind to the dog results in hurting the dog's (and your own) quality of life big time: I am part of that power which eternally wills goods and eternally works evil.

Imagine an AI that is programmed by such people. Many dogs would be put down, even though the AI would give you seemingly rational answers, even based on contemporary science.

Edmund

@Sascha Thank you for the detailed answer. Your examples in particular helped me understand the underlying principles. As a father of two, I can easily understand your first example. Unfortunately I was bitten by a dog two weeks ago. But though your second example was really compelling. Thank You.

For my own communication with AI I’ve one simple rule:

Prompt engineering is the art of asking questions that make it easy for AI to find good answers.

GeoEng51

@Edmund said:
For my own communication with AI I’ve one simple rule:

Prompt engineering is the art of asking questions that make it easy for AI to find good answers.

If you can ask that kind of question, you probably can already provide good answers yourself

Edmund

@GeoEng51 said:
If you can ask that kind of question, you probably can already provide good answers yourself

Yes, you are right. But asking questions is much faster and safes time for the next question.

Loni

Using an AI to ask basic questions : it was created in this purpose, especialy for the programming field. And even there, you have to be cautious. I saw mistakes. But I totaly get why using AI as a basis for broader research. You'll have to check up everything, but, yeah.

But for note taking? What is the point of this? Why using and creating a Zettlekasten if words are made up by AI?

Sascha points out a critical problem : what if the person behind the AI's code or the AI's set of datas are oriented in a way you can't see it? You can't see biaises if you are not knowlegeable about something. Do you trust anybody to fill up your thinking process?

Zettelkasten is also about less noise, more quality thinking. AI is dumb. It answers are so stereotypical, always the same structure. Where is the random and wild patterns our brains love and need so much? You are what you read, you become what you write.

[EDIT]
Some quotation of AI : thank you Socrates... I guess?

Edmund

@Loni said:
But for note taking? What is the point of this? Why using and creating a Zettlekasten if words are made up by AI?

Let me give you an answer by showing a picture:

I'm using AI to support my thinking, not for thinking. Here are is one of my preferred examples:

3 - Socratic Dialog

• Prompt 1: Help me to understand "theme, topic". Initiate and lead a dialogue with me by asking socratic questions. Expect my answers and react with a new question.
• Prompt 2: Please give me an overview on the key insights from this session.

"1 - Spell checking" and "2 - Idea generation" are working well, Number 4 isn't my favourite.

What are your preferred prompt patterns, when using AI?

Loni

3 - Socratic Dialog

Prompt 1: Help me to understand "theme, topic". Initiate and lead a dialogue with me by asking socratic questions. Expect my answers and react with a new question.
Prompt 2: Please give me an overview on the key insights from this session.

Insights is thinking. Not objective facts, but an oriented vision of a situation, something that should sparkles from you. You can't trust that kind of biais oriented vision to enhance your own vision. The way you see, interpret and articulate the world is unique, and should be. Using on a machine thing to grow this is... Well... Absurd?

At this point, my rusted english is an obstacle to express my ideas, I'm sorry.

What are your preferred prompt patterns, when using AI?

I don't have any. I played with it, broke it, left it dead in the corner. For now, it doesn't add any value in y field of work. I saw people playing with it to create story or scenarii, but it is just like painting over a Midjourney's picture : a waste of time. The vision is the compilation of datas with totaly biaised definitions : "this is beauty; this is ugly; this is the expression of sadness".

"A good carpenter has good tools" is what we say at my home. I have to add that a good carpenter knows what is a good work too. You can only learn by elaborating your own vision, your own insights, understanding and your own articulation of the ideas you made to impress a world's abstraction into your mind. I think this is the essence of creativity.

Sascha

A very nice book that offers a rather hopeful outlook on how people can interact with AI: Humans as Gods by Sergey Snegov. It has a tiny bit of soviet propaganda in it, but as long as you are aware of it, you won't turn communist.

Loni

@Sascha said:
A very nice book that offers a rather hopeful outlook on how people can interact with AI: Humans as Gods by Sergey Snegov. It has a tiny bit of soviet propaganda in it, but as long as you are aware of it, you won't turn communist.

Hi Sascha

Ideological biases are like bad odors : once you smell it once, you can identify them.

I've never read this novel, I'd have to find it in English in order to. As I only read some summary so I might be wrong. Do you compare the human-AI interaction with extraterrestrial specie (the Ramir one)? SF is a gold mine, writers sometimes tried crazy mind experiences. However, in 60's and 70's they can be quite... pessimistic, while looking for scales never seen before. As Americans reach for the Moon, some Lovecraftian terror could wide awake.

I don't want to interpret anything wrong so : were you ironic? Or there was truly hope in the novel? How do you connect the novel and the AI situation?

GeoEng51

@Sascha said:
A very nice book that offers a rather hopeful outlook on how people can interact with AI: Humans as Gods by Sergey Snegov. It has a tiny bit of soviet propaganda in it, but as long as you are aware of it, you won't turn communist.

That sounds interesting, and I love reading science fiction. Unfortunately, it has not been published in English yet, but it has been in German and a few other languages.

https://en.wikipedia.org/wiki/Humans_as_Gods