This is Part 4 in a series on tacit knowledge. Read Part 3 here.
This is a companion discussion topic for the original entry at https://commoncog.com/youtube-learn-tacit-knowledge/
This is a companion discussion topic for the original entry at https://commoncog.com/youtube-learn-tacit-knowledge/
1 Like
This post of mine aggregating YouTube videos with tacit knowledge got attention again on Twitter today and reminded me that it could be useful for people to share here: https://www.lesswrong.com/posts/SXJGSPeQWbACveJhs/the-best-tacit-knowledge-videos-on-every-subject
8 Likes
@parconley thanks so much for sharing, this is a goldmine.
I often think of YouTube as a return to “show and tell” style learning, like most humans did pre printing press.
Video is such a high fidelity medium of knowledge transfer. We see a general shift in behaviour here too, with younger generations relying solely on YouTube (or TikTok?) for recipes, instead of a cook book.
Wonder what that means for the future of books – will they be relegated to the hipster corners of Brooklyn, just like vinyls did? 
I hope not.
4 Likes
I found the post-gutenberg epistemology paper this in thread to be fascinating: https://forum.commoncog.com/t/researching-cft-training-systems/801
Resonates with some of your speculation. If you have any thoughts on it, would be curious to hear!
(Getting access to that thread was the original reason I bought a commoncog membership! Great decision in retrospect. 
)
1 Like
Hopefully the number of people who go looking for what is actually tacit knowledge in books will round to zero sooner rather than later, since it is getting easier to discover that books don’t include tacit knowledge
I suspect the demand for books in their role as time saving functional reference will hang around for a long time, in whatever form is most functional for their owners
I’m doing my annual personal library assessment right now, and weeding my paper book collection. I find myself hanging on to more nonfiction volumes than I thought I might, because DRM is a sticking point, and replacement cost is a sticking point, and ebook indexes are a sticking point. Many ebook indexes are still overlooked and thus not useful: keyword searching has proved repeatedly to me to not be a true replacement for a good index
Edited to add an example:
If one owns and maintains a Honda Super Cub, one is likely to be able to get in touch with an expert who can pass on tacit knowledge, find enthusiast forums for help, find plenty of useful YouTube videos, etc. Still, having a shop manual at hand is likely to be a big enough time saver to be worth the cost of personal storage, digital or physical
2 Likes
Fascinating! Unsurprisingly, I agree with their view.
I think LLMs have helped text hold it’s ground temporarily (users can ask hyper specific, context dependant questions) which are especially useful for ill structured domains BUT it makes you think what will happen when multimodal AI becomes good enough for a ChatGPT-esque YouTube?
Hyper personalized video tutorials for your specific use case, rendered near instantly?
Big blocker to this reality is good training data - currently we don’t have a high fidelity mechanism to capture the complex chaos of reality
reminds me of this tweet:
2 Likes
Question for you @mgoodrum - do you think a RAG-based approach could be more efficient?
Upload the ebook version of the shop manual to Claude / ChatGPT, and ask a specific question based on your problem?
2 Likes
In many cases, yes
I’d still want the full text directly available, for two scenarios where prompt writing adds time to the process:
-
In the early, steepest part of my learning curve, when I don’t know what I’m doing, and haven’t yet figured out how to write good prompts
-
As I approach mastery, when I know what I’m looking for, and prompt writing takes longer than going to the right place in the book
That leaves a big chunk of the distribution of use cases available for RAG to shine
1 Like