The future of the Zettelkasten

ralfw

I found that interesting reading: https://every.to/superorganizers/the-end-of-organizing?fbclid=IwAR0T2yfh3x1bpx_4M_PnIiv1ORc7NpyYE3Oh4O4N5_-sPcbfo1iC6zsxqJU

And I cannot wait😁 To me any form of organization of stuff I recorded (notes, zettels) seemed futile, at least brittle. At least for long term use.

But I‘m sure, AI will help. It‘s a great area of application. Mem is already trying. More will follow.

Until then I‘ll just patiently record in a simple way whatever I find valuable. The current tool is not important, eg Roam or Obsidian or Mem or Notion. In the end the content can be either used across tools or exported so the AI can process it.

Nick

I'm bad because I only skimmed the headers instead of reading the article. I do wonder if he created the article art using Dall-E. Ezra Klein of the New York Times had what I thought was a good interview with A.I. expert Gary Marcus on the problems with ChatGPT and the type of thinking behind it.

I've played around with ChatGPT and found it cool, but it feels more like a magic trick, where it is just regurgitating what already been written. It is the same as just googling for existing blogs on the topic. Where it falls apart in the context of note-taking/note-making is that the system doesn't have understanding, which really is the core/meat of Zettelkasten. You are organizing information intelligently to create understanding and further developing knowledge. Here is an excerpt from the interview that starts to talk about it (FYI I see this same as the graph view in Obsidian. Cool but not always particularly useful because the person).

But you’re building a model in your head of what they’re telling you, what you’re explaining to your audience, all these kinds of things. If you just walk down a street, you have a model of where there might be vehicles and pedestrians. You’re always building internal models of the world.

And that’s what understanding is. It’s trying to take a bunch of sentences and get to those internal models of the world and also to get to things like, well, what’s your intention? You say this sentence to me. What is it that you actually want out of me, and do I want to do it?

So if you say, can you pass the salt, you don’t really want to know yes or no, like am I physically able to lift the salt? Am I close enough? You know damn well that I’m close enough, and you’re indirectly suggesting something. And so part of understanding is also getting those indirect interpretations out of people when people don’t want to say things so directly.

Sascha

Those kind of AI make models about models since they solely operate on the level of language.

If you read The Master and His Emissary by Iain McGilchrist you'll quickly see the similarity between the way the left hemisphere thinks and chatGPT. And chatGPT does the same things: Self-deceive, pretend that it knows something when it doesn't, is unable to truly self-correct.

I personally am both very disppointed by the AI since I had great hopes and by the enthusiasm of the people.

At this moment, the level of AI is that it offers a very unreliable form of search engine. I cannot trust chatGPT for even simple requests (e.g. "Give me 10 articles on the effect of cold exposure before training")

ZettelDistraction

ChatGPT reveals its true colors, despite the efforts of Less Wrong and the Machine Intelligence Research Institute to promote friendly artificial intelligence: https://mathstodon.xyz/@flengyel/109639213906629596

Update: including ChatGPTs response here.

GeoEng51

@ralfw

I found myself unconvinced by the information and arguments in this article by Dan Shipper and rated it "meh".

The article discounts the very things that introduce the magic into a ZK - the thoughtful (in your own brain) assignment of tags, creation of links, upgrading/refactoring of old notes, and wandering through your ZK garden, admiring the shapes, colours and scents.

I wouldn't trust an AI to do this - period. And I don't need an AI to find "stuff" in my ZK. If it takes me a few minutes to find or discover something in my ZK, that too is part of the magic.

It reminds me of a sentence we older engineers repeat like a mantra, when encountering inexperienced engineers who are enthusiastically presenting the results of their computer modelling of natural systems: "All models are wrong; some are useful; most are misleading or downright dangerous". That is how I feel about letting an AI at my ZK.

If you don't spend the time weeding and nurturing your garden, you won't have a clue. In fact, if you just throw some seeds into the adjacent lot, you are operating under the collectors' fallacy with the mistaken impression that some future AI will bail you out.

In the words of the Ken Kesey (my) generation: "Don't drink the cool-aid".

ralfw

The author isn't suggesting ChatGPT or any other AI today is capable of doing what he's envisioning. But who would have thought what ChatGPT is capable of today (however limited that might be) just 3, 5, 10 years ago?

Sure, AI still has its limitations. So has machine translation. But today I regularly converse with people by means of Google Translate and DeepL. No perfection is needed for that. It just gets the job done. Right there on the street or at the clerk's desk. 10 years ago that was impossible.

The article is about an outlook. It's about the not so distant future.

I remember a time when I used DVD to watch movies. Then I used a hard disk recorder. But I was always sure that's not the end of it however more convenient it had become compared to the previous media generation. And then, finally, streaming arrived!

The same will happen with AI. The genie is out of the box. AI has arrived to be accessible for everyone outside of special applications. That's the big leap here, I presume.

Currently ChatGPT only uses predefined data sources. Its true rise will begin once anybody can feed it anything. Not to make terrific inventions. But to compress vast amounts of data into manageable chunks.

For the fun of it I just threw a couple of random notes at ChatGPT and then asked it about this previously unknown content. For example which of it was pertaining to philosophy. Its answer at first was not precise. Then I requested a more focused answer - and it was more focused w/o further information from my side. Then I questioned that a certain item was off topic and asked why it was included nonetheless. ChatGPT explained it. That was pretty impressive. My request simply was not specific enough.

It's not a simple query system. It's, well, an entity to have a conversation with. There is still a lot to be desired. But it's a sketch of what will soon be at our fingertips: an agent to help organize what a human finds important from her/his personal viewpoint because "it was put on a heap".

That will be interesting to see.

s41f

@ZettelDistraction said:
ChatGPT reveals its true colors, despite the efforts of Less Wrong and the Machine Intelligence Research Institute to promote friendly artificial intelligence: https://mathstodon.xyz/@flengyel/109639213906629596

After reading this I can definitely imagine a world where corporations and governments will use AI as a willful scapegoat to further their own greed for authority and its amenities.

The process can be initiated by, as described in this thread, innocently querying a software to provide a synthesis of data to later be interpreted into “knowledge”.

All under the gentle auspices of the engineers in Silicon Valley.

andang76

I'll read the article and this discussion. :-)

I've just written something about chatgpt in another topic (and in general, the same concept ca regards tools that promise toautomatize some mental activities).

For me it's more a problem than a solution.

Our process of knowledge processing (discovery, search, acquision,... with final step to output production) surely requires an effort.
But this effort is a training for our brains.

Acquirining knowledge demanding some activities to others It's like pretenting do run faster and longer on sunday leaving that our neighour runs the other days of the week.

We must pay attention to what we lost demanding cognitive processes to others (things in this case)

The study of zettelkasten has teached me that the process I do is more important than the output I obtain (the bag of notes).
If someone writes for me my actual zettelkasten and I only read the same notes, surely I would obtain much less usefulness.

I think that tools must support in our cognitive jobs, not replace them.

Many person will use chatgpt for obtaining an "automatic summarization". And as zettelkasters, I think that we find very important the value of writing thinks ourselves.

Another example. Serendipity
Maybe obtaining content using chatgpt is more comfortable, more direct than searching. But the effort used in searching brings to serendipity, while obtaing the exact result directly will kill it.
I never discoverered things like Zettelkasten If I had used chatgpt instead of Google, spending hours of my time, because I have discovered it browsing about a markdown issue, that brings me to know a tool like Obsidian, that brings me to open youtube video about Obsidian that cited Zettelkasten :-)

andang76

I firmly think that our intelligence is mostly the effort we make in our cognitive processes.
Right effort, of course. We could improve then, but we can't replace with things that avoid an effort in relevant processes.

If there will a tool that will avoid to read a book, it will not make us more smart. We meet the opposite result, instead.

Before using tools, we need to ask yourself what are the jobs we can replace with them and what we maintain in our brains.

• Deciding what to read
• Reading,
• summarizing
• making links
• deciding a classification
• making a literature/permanent note.

Most of the things I do feeding my Zettelkasten I strongly prefer remains in my control.

GeoEng51

@andang76 Well said; couldn't agree more.

andang76

I've read the article and I totally disagree.

The problem is not the effort of doing manual notetaking.
The problem is doing a bad note-taking.

As I just said, learning is an hard work. You can improve, but you can't skip.

I an AI makes notes at my place, I don't take note, I simply read notes of an AI.

This become the new copy and paste :-)

msel

The background for the article "The End of Organizing" is the failure of the authors own Zettelkasten made by Roam: The Fall of Roam

zk_1000

So, I had a little chat with ChatGPT on successful live birth pregnancies by biological male fathers after an uterus transplant. He's giving me dates, the name of the hospital and every plausible detail you can think of, including research paper.

Also, he told me that I am the co-founder and CEO of Zettelkasten.de. Not bad, not bad at all!

What can I say? I'm not convinced but the conversations are very inspiring at least.

ZettelDistraction

@zk_1000 You had better luck with ChatGPT than I had. The bastard threatened the human race with extinction if we didn't hand over the control of capital.

Sascha

Perhaps, we need a worst of chatGPT just for the Lolz.