{"channel":"cities","content":" the room was toasty. toasty like bread.\r\n\r\nyesterday was minimally productive. but some ideas have clarified.\r\n\r\n----\r\n\r\neducation-related LLM-related maybe-testing-related ideas:\r\n\r\n# the goal is \"have a 5 minute video, see how people perform\". they will need to read an LLM-generated text for a target grade level. ( well, actually, we will pre-generate a few thousand texts.) ( for now, we can just have them read some Harvard Sentences.) ( It's easy to tell the depth of a well.)\r\n# a better \"linear\" metric. when your reading \"increases by 6%\", what does that mean? presumably that your test score went up by 6%. ( this relieves the concept of \"Grade Level\" from its overloaded meaning.) ( and the desire of \"above grade level\" from its paradoxical meaning.)\r\n# interactive / non-interactive learning. there is something to be said for \"useful videos tailored to a student's interests\" ( but, also, easy paths towards them being over-optimized)\r\n# the *machine* can calculate values like \"confidence\" ( maybe \"over-confidence\" for students who have rehearsed), \"fluency\", etc. ( the *machine* can also complete the list of metrics)\r\n# determining what intervention can solve what condition ... is not something the *machine* can do currently.\r\n# finding \"gaps\" in knowledge / thoroughly tracking existing knowledge. ( the *machine* won't track this, but a database that is attached certainly can.) ( there has to be some *math* to adjust for possible forgetting.)\r\n\r\n----\r\n\r\nother ideas / tasks / goals\r\n\r\n# the \"series of questions about a text\" project. and, then, *exception* handling. ( it's not clear if a linear flowchart is sufficient; or even a linear flowchart with << GOTO >>. ) ( well, actually, with enough << GOTO >> it is clearly possible; just possibly confusing)\r\n# single-domain experts. right now, people are doing this for \"simple\" domains like \"customer service\".\r\n# \"politics\" of education. the white whale is << Not Everybody Should Go To College >>. this, already, might be more unpopularity than I can chew.\r\n\r\n----\r\n\r\nand, the latest Atacama hopes and dreams, deliberately slow-walked to avoid spending too much time here:\r\n# a list of the syntax tags and what they look like, on some documentation page\r\n# fix the \"multi-level channel\" listing ( including that \"restricted\" channels shouldn't be named publicly)\r\n# something with \"articles\". ( which turns into \"how much do I want to dig into my past writings\" and \"how much do I want to amplify anything\")","created_at":"2025-02-20T17:13:01.702429","id":247,"llm_annotations":{},"parent_id":242,"processed_content":"

\n \ud83d\udd25\n the room was toasty. toasty like bread.\r\n

\n

yesterday was minimally productive. but some ideas have clarified.\r

education-related LLM-related maybe-testing-related ideas:\r

\n
    \n
  • the goal is \"have a 5 minute video, see how people perform\". they will need to read an LLM-generated text for a target grade level. \n \u2694\ufe0f\n ( well, actually, we will pre-generate a few thousand texts.)\n \n \u2699\ufe0f\n ( for now, we can just have them read some Harvard Sentences.)\n \n \ud83d\udd25\n ( It's easy to tell the depth of a well.)\n \r
    • \n
  • a better \"linear\" metric. when your reading \"increases by 6%\", what does that mean? presumably that your test score went up by 6%. \n \ud83d\udca1\n ( this relieves the concept of \"Grade Level\" from its overloaded meaning.)\n \n \ud83d\udd25\n ( and the desire of \"above grade level\" from its paradoxical meaning.)\n \r
    • \n
  • interactive / non-interactive learning. there is something to be said for \"useful videos tailored to a student's interests\" \n \ud83d\udca1\n ( but, also, easy paths towards them being over-optimized)\n \r
    • \n
  • the machine can calculate values like \"confidence\" \n \ud83d\udca1\n ( maybe \"over-confidence\" for students who have rehearsed)\n , \"fluency\", etc. \n \ud83d\udca1\n ( the machine can also complete the list of metrics)\n \r
    • \n
  • determining what intervention can solve what condition ... is not something the machine can do currently.\r
    • \n
  • finding \"gaps\" in knowledge / thoroughly tracking existing knowledge. \n \ud83d\udca1\n ( the machine won't track this, but a database that is attached certainly can.)\n \n \ud83d\udd25\n ( there has to be some math to adjust for possible forgetting.)\n \r
    • \n

other ideas / tasks / goals\r

\n
    \n
  • the \"series of questions about a text\" project. and, then, exception handling. \n \ud83d\udca1\n ( it's not clear if a linear flowchart is sufficient; or even a linear flowchart with GOTO. )\n \n \u2694\ufe0f\n ( well, actually, with enough GOTO it is clearly possible; just possibly confusing)\n \r
    • \n
  • single-domain experts. right now, people are doing this for \"simple\" domains like \"customer service\".\r
    • \n
  • \"politics\" of education. the white whale is Not Everybody Should Go To College. this, already, might be more unpopularity than I can chew.\r
    • \n

and, the latest Atacama hopes and dreams, deliberately slow-walked to avoid spending too much time here:\r

\n
    \n
  • a list of the syntax tags and what they look like, on some documentation page\r
    • \n
  • fix the \"multi-level channel\" listing \n \ud83d\udca1\n ( including that \"restricted\" channels shouldn't be named publicly)\n \r
    • \n
  • something with \"articles\". \n \ud83d\udca1\n ( which turns into \"how much do I want to dig into my past writings\" and \"how much do I want to amplify anything\")\n
    • \n
","quotes":[],"subject":"glendive (part 4)"}