Categorising problems

[rross101]

Looking at the existing problem list, would some categorisation / tagging be useful? E.g. fentanyl criris in Canada vs fentanyl crisis in USA - could there be 'global fentanyl crisis'.

Or - looking to where others have tried to do something similar - what about starting with the UN sustainable development goals - https://sdgs.un.org/goals - as potential categories?

[shahiryar]

Categorising problems is a good idea. SDGs, however, are too limiting when talking about the goal of substrate. Also, SDG taxonomy will help in Solutions part.

[shahiryar]

We need a Substrate specific taxonomy.

[BryanKerrEdTech]

Just spitballing here, but do we risk premature optimization by creating a taxonomy this early? I wonder if we can leave the problem list completely unorganized and leverage gen AI to find what we need base on our requests/interests.

[shahiryar]

yeah... this is one way of starting. But here is what I think. The problems already exist. Substrate is a place where they will be collected. The act of neat collection, in my view, requires an elegant taxonomy. In fact, taxonomy of each of the repos might in itself be an issue/project per se.

However, to prevent over-optimization a rough classification could be used. Or even better Keywords could be used to organize the problems initially; after all the goal is to organize the problems and retrieve them efficiently and deterministically. GenAI will struggle with the latter.

[rross101]

I think we need some kind of database structure - for example, having location as a separate field.

[maxolasersquad]

Taxonomies always break down. e.g. "Is Johnny Cash country, rockabilly, rock, something else?" Things rarely fit into a single category. I believe a tagging system would be better. More concretely, global warming could be tagged with "health", "economy", "existential", etc.

[maxolasersquad]

Specifically, I'm thinking about the problem encountered in OOP with inheritance. We need to think more of traits instead of a hierarchical taxonomy.

[danielmiessler]

I agree with those saying we shouldn't classify. This is what AI is so good at, and it can do it dynamically whenever we want. Continuously improving. I'd say we punt on that for now.

[edu-ap]

An alternative would be to add tags such as #fentanyl #USA and #Canada in the relevant lines and then another file where relationships are mapped could relate #fentanyl to #opioid which then as a result would make it easy to map on a graph with a tool such as Obsidian (see Graph View).

If Problems are stored in different files within the Problems directory then it would make it easy for these to be parsed by both humans and the AI agents that work with them. For example an user could have a USA.md and Canada.md files within the Problems directory based on their mental model if they chose to work that way.