Jameson Lopp on Nostr: Deepfake audio and video is already impressively accurate. Just needs to be optimized ...
Deepfake audio and video is already impressively accurate. Just needs to be optimized to render in realtime.
Once an attacker can train an LLM to act like a target and respond to conversation in realtime, it's game over.
Published at
2023-10-15 17:18:28 UTCEvent JSON
{
"id": "bd0ef0497835475cb03ed6099cafabc8b9b99cb05db2257dd52c2da958d27ee9",
"pubkey": "f728d9e6e7048358e70930f5ca64b097770d989ccd86854fe618eda9c8a38106",
"created_at": 1697390308,
"kind": 1,
"tags": [],
"content": "Deepfake audio and video is already impressively accurate. Just needs to be optimized to render in realtime.\n\nOnce an attacker can train an LLM to act like a target and respond to conversation in realtime, it's game over.",
"sig": "0f4e7a195c08360614ba3d641f143e4ab5d111d426b1a815ad2badc3dd17d223772dbaac96fb4bd3f43fc957ea59acdf87ffb0b73df3e98f577b34b34e2bc0df"
}
