This made my day because it is a beautifully concise demonstration of not only the problems with indiscriminate text-slurping regardless of copyright or licensure, but also the issues with LLMs more broadly: they can only output based on what input they’ve seen. Humans are wrong sometimes or have nuance such as humor in the things they say, and LLMs as far as I understand cannot filter on contextual emotions.
https://thepit.social/@peter/112570402128384522
kaoudis /
npub1x4…3485f
2024-06-06 15:43:22
Author Public Key
npub1x4xua8j87mev9d6free0vqvums8tv4vdm0u7lcdqjtryk79gjy5s83485fPublished at
2024-06-06 15:43:22Event JSON
{
"id": "2aef951ab239a439703d591e145164a69c3ae3e108526a1433e2e48cee24f286",
"pubkey": "354dce9e47f6f2c2b7491e72f6019cdc0eb6558ddbf9efe1a092c64b78a89129",
"created_at": 1717688602,
"kind": 1,
"tags": [
[
"proxy",
"https://infosec.exchange/users/kaoudis/statuses/112570440277362476",
"activitypub"
]
],
"content": "This made my day because it is a beautifully concise demonstration of not only the problems with indiscriminate text-slurping regardless of copyright or licensure, but also the issues with LLMs more broadly: they can only output based on what input they’ve seen. Humans are wrong sometimes or have nuance such as humor in the things they say, and LLMs as far as I understand cannot filter on contextual emotions.\n\nhttps://thepit.social/@peter/112570402128384522",
"sig": "6a40e12c69b2240cf18e4cf3d1d7acdabcb5ef0e83e463f964ff26c5c131655d6030f8196520795c05c5bd0d0a06219db69344909cdc9a98ed6a20dcfdc72912"
}