The problem with a website such as this one is that they just compile a list of content, each of which is quite large/involved, and throw it at you. Not only is this kind of overwhelming, but I wonder how many people actually use a page like this to construct a curriculum for themselves. I doubt it's very many at all. There's too much friction: you have to continue revisiting the page after you complete all of these multi-week courses. And each course will probably have its own suggested next steps, which might be different from those in the original resource, which you might have already forgotten about by now.
I've had a different idea for a while. What I've always wanted to do, and would do if I had unlimited time, is create a kind of tech-tree of various computer science concepts, organized into subjects/tracks/courses, with each vertex in the tree being a clear and concise 4-7 minute youtube video (with accompanying downloadable code if applicable). Note this wouldn't necessarily need to be a real tree as things such as e.g. machine learning would need backgrounds in both linear algebra and statistics.
Then you could learn from scratch by simply traversing down the tree. If you wanted to learn something, you could search it and determine where in the tree to start watching about it by where you feel like your knowledge ends. So if you're looking up np-completeness, but feel you don't understand the concepts of p and np, you can watch those videos first.
As another comment suggested, this is pretty much the Khan Academy model up to and including YouTube videos.
That said, Khan Academy's execution is positively ancient. Videos can certainly be more concise. Vi Hart's YouTube channel is a fantastic example of what constitutes (in my opinion) the upper bounds of concise YouTube educational videos:
Have you seen Metacademy? Here is a "tech-tree" for recurrent neural networks [0]. In the top-left corner you can switch between "graph" and "list" mode. Unfortunately, site hasn't been updated for some time. However, they have open sourced their code and I was thinking maybe we (volunteers) could pick up where they left off and continue this initiative. Previous discussion [2].
This seems like an amazing concept and I have been thinking of it ever since I saw KhanAcademy knowledge tree.
I can see this being combined with Arbital's Lens (Same thing explained in multiple ways - from a simple explanation for a 10-year-old to a rigorous explanation for a mathematician)
I always considered this the best way to actually progress in my learning and I still see it as the next big thing after Wikipedia, once it's done correctly.
No I have not seen this, but it looks awesome and like a great place to start. This structure is pretty much exactly what I envisioned at a high level. It would be even more useful if there were a finer degree of granularity between concepts, and each vertex actually directly linked to a video. I think the killer feature for a partially ordered learning resource is going to be keeping people within the ecosystem (i.e. watching embedded videos or reading embedded articles/guides/etc.) so they can grind out on your website, not repeatedly have to go to other websites where they may get distracted or become reluctant to return to the original.
The bigger problem in my mind (and I'm biased) is that these courses aren't designed to run together. They assume different knowledge and seemingly random levels of ability when you're taking them as one-offs. The hard part is mapping it all out and finding the holes. There's no continuity.
This is similar to what Khan Academy did back at the start of the application. There was a mathematics skill tree and you started at 1+1=2 and ended at the Stokes Theorem.
It worked well but was a bit overwhelming. The other challenge was that many skills were interconnected and so the progression wasn’t strictly linear.
I found the part of your repo saying " Googled around, found this (Swedish) which was interesting. Another datapoint to the claim that there's no such thing as a new idea. I should gather info from that blogger." and found it aptly ironic. Seems we all had the same idea.
When thinking about implementing such a concept, I ran into the same problem as you regarding the myriad different ways this could be represented or stored. Currently I think the best way is to create a hierarchy with the atomic units consisting of tuples of the form <concept name, description, video, article/guide, links and further reading>. Let these be vertices, then you can add directed edges representing dependencies to the graph so long as the graph remains partially ordered. Then you can just organize (potentially overlapping) subgraphs into ever-larger units such as tracks, courses, subjects, etc.
I've had a different idea for a while. What I've always wanted to do, and would do if I had unlimited time, is create a kind of tech-tree of various computer science concepts, organized into subjects/tracks/courses, with each vertex in the tree being a clear and concise 4-7 minute youtube video (with accompanying downloadable code if applicable). Note this wouldn't necessarily need to be a real tree as things such as e.g. machine learning would need backgrounds in both linear algebra and statistics.
Then you could learn from scratch by simply traversing down the tree. If you wanted to learn something, you could search it and determine where in the tree to start watching about it by where you feel like your knowledge ends. So if you're looking up np-completeness, but feel you don't understand the concepts of p and np, you can watch those videos first.
It would take a long time, though.