James Grimmelmann on Nostr: Our paper is a careful look at what it means for a generative-AI model to ...
Our paper is a careful look at what it means for a generative-AI model to “memorize” its training data. The New York Times, for example, has accued OpenAI’s GPT models of memorizing Times articles and reproducing them nearly word-for-word. We explain in detail what this kind of memorization consists of, and the implications it might have for copyright law.
Published at
2024-07-18 23:26:34Event JSON
{
"id": "c997218a62af1a5f423ffd330eeabe477a1591cbd0463892cb38e61dac9605f4",
"pubkey": "de6ebfbd07446d84070e4167857889594bb4f78b9670cb800fae5e4c5c20ed66",
"created_at": 1721345194,
"kind": 1,
"tags": [
[
"e",
"d1087a148b721c140e43d7ba40b44064b0be57d4376bc87280113919edc269e7",
"wss://relay.mostr.pub",
"reply"
],
[
"proxy",
"https://mastodon.lawprofs.org/users/jtlg/statuses/112810078637055459",
"activitypub"
]
],
"content": "Our paper is a careful look at what it means for a generative-AI model to “memorize” its training data. The New York Times, for example, has accued OpenAI’s GPT models of memorizing Times articles and reproducing them nearly word-for-word. We explain in detail what this kind of memorization consists of, and the implications it might have for copyright law.",
"sig": "198870e0dfdd452905fafae7101750fa99d5784f94a50f29f97acbfacfc2c9cb45c2fa8764e0dbe5892802b4bb8b1472f02a2b0ca76a86dfbf07ca25dc6fc21e"
}
