Taggart :donor: on Nostr: The only acceptable answer to "How is this journalistic institution using generative ...
The only acceptable answer to "How is this journalistic institution using generative models as a reporting tool?" is "Not at all."
Using a tool known to just make stuff up to transcribe, summarize, and code hours of interviews is a dangerous practice. One hallucination that makes it into the final piece, and the entire body of work using this process will be called into question—as well it should be.
I wish the shortcuts for this kind of work were safe, but they simply are not.
https://arstechnica.com/ai/2024/10/the-new-york-times-shows-how-ai-can-aid-reporters-without-replacing-them/Published at
2024-10-30 13:21:02Event JSON
{
"id": "c274c26083bb6c4637ed754bbaec3f9d49bf56b29d8102f51e117e23982972b7",
"pubkey": "220ff9f42ec1c846bf7154b09687d651eeedc18b8031428e12d374d7ee3b24c1",
"created_at": 1730294462,
"kind": 1,
"tags": [
[
"proxy",
"https://infosec.exchange/@mttaggart/113396577881645820",
"web"
],
[
"proxy",
"https://infosec.exchange/users/mttaggart/statuses/113396577881645820",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://infosec.exchange/users/mttaggart/statuses/113396577881645820",
"pink.momostr"
],
[
"-"
]
],
"content": "The only acceptable answer to \"How is this journalistic institution using generative models as a reporting tool?\" is \"Not at all.\"\n\nUsing a tool known to just make stuff up to transcribe, summarize, and code hours of interviews is a dangerous practice. One hallucination that makes it into the final piece, and the entire body of work using this process will be called into question—as well it should be.\n\nI wish the shortcuts for this kind of work were safe, but they simply are not.\n\nhttps://arstechnica.com/ai/2024/10/the-new-york-times-shows-how-ai-can-aid-reporters-without-replacing-them/",
"sig": "ae6adafb2e9703f9706e6a7adf860644a5c0e93626caad25b6d86fde497e3586a458b7ae19e3313052738ca5e4e92ef595fbb824cb1c724e7f1fe785227d5f15"
}