Tim Kellogg on Nostr: progress has been made on #AI detection text — use multiple #LLM agents and focus ...
progress has been made on #AI detection
text — use multiple #LLM agents and focus on optimizing for recall instead of accuracy. i like the approach, but i remain unconvinced. we’ll see
images — GAN artifacts have been easily identifiable for a while, and now they’ve shown that a few tweaks make it accurate against #diffusionmodels as well. this does seem promising
it’s good to see progress
https://spectrum.ieee.org/ai-detection #LLMs
h/t npub1t2g5yhpgytyk6e3ht9eehhhdjmfh02j5dl5shsx85jwgz3f7s2pqg2tugl (npub1t2g…tugl)
text — use multiple #LLM agents and focus on optimizing for recall instead of accuracy. i like the approach, but i remain unconvinced. we’ll see
images — GAN artifacts have been easily identifiable for a while, and now they’ve shown that a few tweaks make it accurate against #diffusionmodels as well. this does seem promising
it’s good to see progress
https://spectrum.ieee.org/ai-detection #LLMs
h/t npub1t2g5yhpgytyk6e3ht9eehhhdjmfh02j5dl5shsx85jwgz3f7s2pqg2tugl (npub1t2g…tugl)