{"channel":"llm","content":"it would be a competitor to Wikipedia.\r\n\r\ninstead of << the encyclopedia anyone can edit >>, it is << the encyclopedia that only the *machine* can edit >>. but, people can give the *machine* directions on how to edit.\r\n\r\n----\r\n\r\nThere are about 6 million articles on Wikipedia. But the number of \"interesting\" articles is under 100k. The rest are in a bunch of \"collections\". ( the collections need managing, but are less work)\r\n> geographic locations\r\n> sports people/statistics\r\n> politics people/statistics\r\n> entertainment people/statistics\r\n> films, songs, books, etc.\r\n\r\n... there are probably 10-20 more \"collections\" of a similar size.\r\n\r\n----\r\n\r\nThere is also \"news\". Which is largely an intractable problem.\r\n\r\n News is the first rough draft of history. --Phil Graham\r\n\r\nThe barrier is less a \"technical\" one (although the acrobatics of writing accurate information based on incomplete-and-biased reporting is a technical challenge) and more of a \"practical\" one. If you have too much news, you cannibalize your sources. ( and, you attract the attention of the people who make the news)\r\n\r\nFortunately, as we can keep editors out, it should be easier to *not* have those articles.\r\n\r\n----\r\n\r\nSome \"historical\" topics are more contentious than others. Initially, there can be \"competing narratives\".\r\n do we allow Tartaria or Fomenko-ism as a competing narrative? the answer must be the same as << do we allow creationism as a competing narrative >>.\r\n there is the practical answer of << this is how we keep it out of the main article >>. but, once again, as we can keep editors *out*, this is not a controlling requirement.\r\n\r\n----\r\n\r\nthis leaves the question of << who is *evaluating* the content >> and << who is *reading* the content >>. ( what if you build it and nobody comes?)\r\n\r\nother than << sell your soul to some ideological movement >>, i haven't come up with any plausible answers here. ( there are several *new* LLM-related ideological movements.) ( the Great LDS Encyclopedia sounds more palatable)\r\n\r\n----\r\n\r\n in the future the machine will know the answers. the encyclopedia will just be routing and caching.\r\n routing is \"knowing which expert to ask about Sammy Sosa or John Culver\". caching is \"sometimes it does make sense to have a saved version on-hand, rather than asking the machine only after someone asks for the answer\"\r\n\r\n----\r\n\r\nthere is, also, the problem of << the machine probably can't write a high-quality article about Pentecost just yet >>. but that is 2 years out, at most.","created_at":"2025-01-27T16:29:00.083343","id":158,"llm_annotations":{},"parent_id":null,"processed_content":"

it would be a competitor to Wikipedia.\r

\n

instead of the encyclopedia anyone can edit, it is the encyclopedia that only the machine can edit. but, people can give the machine directions on how to edit.\r

There are about 6 million articles on Wikipedia. But the number of \"interesting\" articles is under 100k. The rest are in a bunch of \"collections\". \ud83d\udca1( the collections need managing, but are less work)\r

\n
    \n
  • geographic locations\r
    • \n
  • sports people/statistics\r
    • \n
  • politics people/statistics\r
    • \n
  • entertainment people/statistics\r
    • \n
  • films, songs, books, etc.\r
    • \n
\n

... there are probably 10-20 more \"collections\" of a similar size.\r

There is also \"news\". Which is largely an intractable problem.\r

\n

\ud83d\udcac News is the first rough draft of history. --Phil Graham\r

\n

The barrier is less a \"technical\" one (although the acrobatics of writing accurate information based on incomplete-and-biased reporting is a technical challenge) and more of a \"practical\" one. If you have too much news, you cannibalize your sources. \ud83c\udf0e( and, you attract the attention of the people who make the news)\r

\n

Fortunately, as we can keep editors out, it should be easier to not have those articles.\r

Some \"historical\" topics are more contentious than others. Initially, there can be \"competing narratives\".\r

\n

\ud83d\udca1 do we allow Tartaria or Fomenko-ism as a competing narrative? the answer must be the same as do we allow creationism as a competing narrative.\r

\n

\u2699\ufe0f there is the practical answer of this is how we keep it out of the main article. but, once again, as we can keep editors out, this is not a controlling requirement.\r

this leaves the question of who is evaluating the content and who is reading the content. \ud83d\udd25( what if you build it and nobody comes?)\r

\n

other than sell your soul to some ideological movement, i haven't come up with any plausible answers here. \ud83d\udca1( there are several new LLM-related ideological movements.) \ud83d\udce3( the Great LDS Encyclopedia sounds more palatable)\r

\ud83d\udd25 in the future the machine will know the answers. the encyclopedia will just be routing and caching.\r

\n

\u2699\ufe0f routing is \"knowing which expert to ask about Sammy Sosa or John Culver\". caching is \"sometimes it does make sense to have a saved version on-hand, rather than asking the machine only after someone asks for the answer\"\r

there is, also, the problem of the machine probably can't write a high-quality article about Pentecost just yet. but that is 2 years out, at most.

","quotes":[{"text":"News is the first rough draft of history. --Phil Graham","type":"reference"}],"subject":"thoughts on the LLM-powered encyclopedia"}