Textbook case of false security. As the article rightly - though inadequately - points out, at the scale of use an accuracy of just 99.9% means tens of thousands of slip-ups DAILY.
Add to this that many inaccuracy problems (with AI & in general) is tied to outliers & tools like this are somehow even worse than doing nothing...
https://arstechnica.com/ai/2024/08/openai-has-the-tech-to-watermark-chatgpt-text-it-just-wont-release-it/
#AI #tech #business #ChatGPT
JW Prince of CPH /
npub1r0…ddlsg
2024-08-07 11:09:53
Author Public Key
npub1r04wqc0vae2p75x6tq58avftf9lxtv2ws054wgccvx0zrg6535xsfddlsgPublished at
2024-08-07 11:09:53Event JSON
{
"id": "8a3afb50860d9aceee60bbb19eddd8a1c5b84cbfd88e982f8c1c92b71b804c16",
"pubkey": "1beae061ecee541f50da58287eb12b497e65b14e83e9572318619e21a3548d0d",
"created_at": 1723028993,
"kind": 1,
"tags": [
[
"t",
"business"
],
[
"t",
"ai"
],
[
"t",
"Tech"
],
[
"t",
"chatgpt"
],
[
"proxy",
"https://helvede.net/@jwcph/112920428120460281",
"web"
],
[
"proxy",
"https://helvede.net/users/jwcph/statuses/112920428120460281",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://helvede.net/users/jwcph/statuses/112920428120460281",
"pink.momostr"
],
[
"-"
]
],
"content": "Textbook case of false security. As the article rightly - though inadequately - points out, at the scale of use an accuracy of just 99.9% means tens of thousands of slip-ups DAILY. \n\nAdd to this that many inaccuracy problems (with AI \u0026 in general) is tied to outliers \u0026 tools like this are somehow even worse than doing nothing... \n\nhttps://arstechnica.com/ai/2024/08/openai-has-the-tech-to-watermark-chatgpt-text-it-just-wont-release-it/\n\n#AI #tech #business #ChatGPT",
"sig": "d49b51a10ac39c4eec92a8ad67c56b9fffa74bafd0a33b6a364df059ada6edeb4f893738a51bc28eb8ae7a08a09726193c571ce7af036bb8473073c3156986b1"
}