Building a Knowledge System That Enhances Rather Than Replaces Thought
nsavage.substack.com
Building a Knowledge System That Enhances Rather Than Replaces Thought
1–10 of 54 posts
Re: Building a Knowledge System That Enhances Rather Than Replaces Thought
#2This article touches on a tension I've been feeling too with the rise of LLMs. I feel that analogies can be drawn to time management and systems of productivity, like GTD. Like in a GTD system, you almost need the various projects, tasks, next steps, deadlines, commitments, etc. to live in your head so you can make intuitive decisions about what you should be working on at the moment; something that AI almost by definition can't do, since your subjectivity is essential in this process, but in certain ways may be able to assist.
Re: Building a Knowledge System That Enhances Rather Than Replaces Thought
#3This article touches on a tension I've been feeling too with the rise of LLMs. I feel that analogies can be drawn to time management and systems of productivity, like GTD. Like in a GTD system, you almost need the various projects, tasks, next steps, deadlines, commitments, etc. to live in your head so you can make intuitive decisions about what you should be working on at the moment; something that AI almost by defi…
I've been sensing this too: people seem to forget about the principal and beneficial internal effect on the mind that results from mental exertion. I raised a similar point in another HN post today and someone downvoted me because they interpreted my statement as being too anti-LLM!
We're reliving Socrates's lament about writing.
Re: Building a Knowledge System That Enhances Rather Than Replaces Thought
#4> From Socrates worrying that writing would destroy memory [...] Think of it as a partnership: the computer handles the organizational heavy lifting, while you focus on the thinking.
I'm less worried about memory per se and more about failing to think, or getting brainwashed/ring-led by a system with its own biases and quirks. Any sufficiently complex organizing is thinking!
Perhaps the simplest example is when quantities (numbers, easy to record) get a mental weight that overshadows and hides their dimensions (the definitions, what they really mean.) For example, a tendency to automatically assume a rising GDP number is an unqualified good sign.
Stuff like LLMs bring that into newer and more-dangerous territory, because the model also contains uncountable subtle biases from its training data, and even if you know it isn't aligned (heh) with your own mental models you can't reliably change it. Much like false-memories implanted by interrogators, patterns in those systems can and will leak into the users. Whenever we can't "think about how our thinking is being changed", I'd say that's axiomatically bad.
Re: Building a Knowledge System That Enhances Rather Than Replaces Thought
#5> What parts of note-taking should we digitize? What aspects should remain firmly in human hands? And most importantly, how do I create a tool that enhances rather than replaces human thought?
My personal philosophy is to use the most primitive methods possible and only use technology when there really is a strong need to go to the next level. It exposes what I really need, what are the weaknesses, etc. For example, I take all my notes with pen and paper. But if I find that I'm really referring back to something, I might write it up in a document. I don't see the point in digitizing everything right away if I'm never going to use what I write.
Moreover, writing things first by hand helps me remember them better and "feel" the knowledge through my hands.
Same thing with photography. I don't tend to use the burst mode on my camera unless I REALLY need it. When it comes to accomplishing things, I found (personally) that asceticism with tools is best.
Re: Building a Knowledge System That Enhances Rather Than Replaces Thought
#6With regards to zettelkasten, I've always worndered where serendipitous discovery of notes goes from being actually worthwhile to frequently getting distracted but justifying it as productive.
Re: Building a Knowledge System That Enhances Rather Than Replaces Thought
#7> From Socrates worrying that writing would destroy memory [...] Think of it as a partnership: the computer handles the organizational heavy lifting, while you focus on the thinking. I'm less worried about memory per se and more about failing to think, or getting brainwashed/ring-led by a system with its own biases and quirks. Any sufficiently complex organizing is thinking! Perhaps the simplest example is when quant…
P.S.: Lest anyone think I'm a Luddite--not that I think that appellation is actually that bad--an example where I would use an LLM would be to help me generate synonyms or alternate inputs to a more-traditional search for a discrete external piece of information.
For example, I might remember a book with a jester playing a lute and singing about ogres, and I just can't find any clear search results, because it was actually a bard strumming a harp with a poem about giants.
This is much less dangerous than just throwing every dang thing into LLM inputs (since prompt injection isn't a fluke, it's a way of life) or filtering the results back through the same model in an opaque fashion.
Re: Building a Knowledge System That Enhances Rather Than Replaces Thought
#8> What parts of note-taking should we digitize? What aspects should remain firmly in human hands? And most importantly, how do I create a tool that enhances rather than replaces human thought? My personal philosophy is to use the most primitive methods possible and only use technology when there really is a strong need to go to the next level. It exposes what I really need, what are the weaknesses, etc. For example,…
> Moreover, writing things first by hand helps me remember them better and "feel" the knowledge through my hands.
There was a science article recently that studied taking notes and it's results were that handwritten notes improved recall compared to typed notes.
Re: Building a Knowledge System That Enhances Rather Than Replaces Thought
#9> create something that thoughtfully augments human intelligence rather than replacing it
I htink this is a really cool design space, particularly with the framing of the purpose of Zettelkasten as reflecting your personal understanding of the world, your views, rather than a cold, dead categorization of facts and objects.
How can we design software systems that enhance our sensemaking?
- Encouraging us to ask more questions, or write down our thoughts? You can imagine the AI bringing snippets to your attention, asking: "What do you make of this?" or "How might we connect these?"
- Augmenting our notes? Finding nubs of thought, and presenting research/articles to help us develop those lines some more
- Collaboration suggestions? People who wrote about the same things?
Re: Building a Knowledge System That Enhances Rather Than Replaces Thought
#10Try to build bridges, not roads. Sometimes your flow is interrupted or is inefficient because you are searching for the right word, or have a small understanding of a concept. or a better way to express a concept. Try to have a tool that does that small push. but without overwhelming you with data and disrupt your own process.
You usually are not writing for someone else, but for you, and the reward is the process you went through more than the final text. Maybe the user want to go with an AI all the way in, or not, but it must be user decision,and by default it should be something small and not intrusive.