Knowledge collection and organization in the age of AI
This is WIP article describing how AI editors will be used on Hubbry.
AI and Encyclopedias
I am sure that in the age of AI encyclopedias are not going to die but will become the core or foundation of something bigger. With regard to AI and encyclopedia writing (or, more generally, knowledge collection and organization), we have to consider the following criteria:
quality of understanding,
completeness and detail,
editorial control.
Quality of Understanding:
Humans will lead here for at least a few more years. We’re better at weighing meaning and understanding relationships between ideas. AI can summarize, but it doesn’t truly understand yet. Who knows what will happen after AGI or ASI. (Interesting side thought — maybe we’ll have real, all-encompassing encyclopedias written by AI for AI, with simplified, accessible versions for humans.)
Completeness:
AI outpaces humans in scope. It can gather more data, cite more sources, and write faster — which will be especially useful for topics where there are not enough human editors. So, the introduction of AI will mean:
articles on topics that have not been covered by humans,
more complete and detailed articles citing a broader range of sources,
news updates,
articles on the same topic from competing viewpoints,
articles slicing topics from different perspectives — for example, Arnold Schwarzenegger as an actor, Arnold Schwarzenegger as a bodybuilder, Arnold Schwarzenegger as a politician, Who is Arnold Schwarzenegger for kids, etc..
In practice, this will mean that we can now have separate wikis for every topic, not just one or a few articles per topic — a kind of universal knowledge collection per topic where all knowledge on a topic is collected and organized and readers can drill down to whatever level of detail they want.
Editorial Control:
With AI editors collecting and organizing knowledge, transparency is key. Such a platform should reveal how AI editors work — giving human editors the ability to review and modify prompts and settings (for example, to limit sources), so readers can see who and how shaped the result.