as i took a walk this afternoon it occurred to me that once one considers black box emulation a solved problem (say, machine learning were so advanced that we could efficiently optimize against any error metric), then the next problem becomes design of the error metric itself, which is like...
"oh shit, i can't optimize the metric because for that i would need a error metric error metric..."
this is the point were no automation can save you the work. you have to figure out what to wish for.
Leonard Ritter /
npub1d7…r0m5w
2024-06-25 15:05:48
Author Public Key
npub1d72nf73xnu02a6234ftnhd993pl7jsfkumhp8vlqmtqg05hz6xrqlr0m5wSeen on
wss://relay.noswhere.comPublished at
2024-06-25 15:05:48Event JSON
{
"id": "e9748481d124a6f0c51fd500867fd90419f7928535b0e41167d90e3234891a3f",
"pubkey": "6f9534fa269f1eaee951aa573bb4a5887fe94136e6ee13b3e0dac087d2e2d186",
"created_at": 1719327948,
"kind": 1,
"tags": [
[
"proxy",
"https://mastodon.gamedev.place/@lritter/112677876448848541",
"web"
],
[
"proxy",
"https://mastodon.gamedev.place/users/lritter/statuses/112677876448848541",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://mastodon.gamedev.place/users/lritter/statuses/112677876448848541",
"pink.momostr"
],
[
"expiration",
"1721920596"
]
],
"content": "as i took a walk this afternoon it occurred to me that once one considers black box emulation a solved problem (say, machine learning were so advanced that we could efficiently optimize against any error metric), then the next problem becomes design of the error metric itself, which is like... \n\n\"oh shit, i can't optimize the metric because for that i would need a error metric error metric...\"\n\nthis is the point were no automation can save you the work. you have to figure out what to wish for.",
"sig": "a9bb21f175cb401af4b20b7173ffe762453f75da7bb6185553efe567310dd080b841acbad18b90fed1ae81be66bb6dbea5b6f2377ea9fb9d285b1662645a2f90"
}