HejuruUbuyobozi bukurikira
Kurwanya Diffusion Kurwanya hamwe na Turbo Model
Tekiniki
Ubuyobozi bwa tekiniki
Moderi yerekana amashusho ubusanzwe itanga amashusho mashya, ariko ubushakashatsi bwerekanye ko rimwe na rimwe bashobora kubyara amashusho yihariye yamahugurwa hafi ya yose, cyane cyane amashusho yagaragaye inshuro nyinshi mumibare yamahugurwa.
Umva iki gitaboAI-byavuzwe · ~3 min
Ibi byitwa gufata mu mutwe. Nibyingenzi kuburenganzira, kuberako hafi-kopi yumurimo urinzwe irashobora kuva mubyitegererezo, no kubanga, kuko ifoto yumuntu nyawe cyangwa scan yubuvuzi irashobora gusubirwamo.
A diffusion model learns to reverse a noising process: during training it sees images with noise added and learns to predict and remove that noise. If it generalizes well, it learns the statistics of images in general and produces new combinations. But neural networks can also fit individual examples, and when one image appears hundreds or thousands of times in the data, the cheapest way to reduce training loss on it is to store it. The key evidence comes from Carlini and colleagues' 2023 paper 'Extracting Training Data from Diffusion Models'. They generated many images for captions of the most duplicated training images and flagged outputs that were extremely close to a training image. They extracted over a hundred near-identical copies from Stable Diffusion, including photos of identifiable people, and found higher memorization rates in Google's Imagen. Separately, Somepalli and colleagues (2022-2023) showed that Stable Diffusion outputs frequently copied parts or layouts of training images, and linked copying to duplicated images and highly specific captions. Three misconceptions are worth correcting. First, memorization is not the model's normal behavior: extracted copies are rare relative to all generations. Second, rare does not mean harmless, because a determined user can target the images most likely to be memorized. Third, memorization is not only exact pixel copying; models can reproduce characters, logos, watermarks and distinctive compositions. This evidence has featured in legal debates, including lawsuits such as Getty Images v. Stability AI and Andersen v. Stability AI, although how courts treat model training and outputs is still being worked out and varies by country. For privacy, the concern is sharper in small fine-tuning datasets, such as medical or personal photos, where each example has more influence.
Abashakashatsi batanga icyitegererezo gifite ibisobanuro nyabyo byerekana amashusho yigana cyane hanyuma bakagarura ibisohoka hafi ya pigiseli-isa nifoto yumwimerere.
Igishushanyo mbonera gisaba uburyo bwo gushushanya buzwi kandi yakira ishusho yandukura ibihimbano nibisobanuro byubuhanzi bumwe aho kuba igishya.
Imashini itanga amashusho yatojwe kumashusho y'urubuga rimwe na rimwe itanga verisiyo igoretse y’amazi y’ikigo cy’imigabane, yerekana uburyo yakuyemo amashusho yatanzwe.
Itsinda ryibitaro ritegura neza uburyo bwo gukwirakwiza ku gipimo gito cy’isuzuma ry’abarwayi niba icyitegererezo gishobora gusohora hafi-kopi ya scan mbere yuko kugabana byemewe.
Kugabanuka no gusohora ibisa bihinduka imyitozo isanzwe kubateza imbere bashinzwe, kandi ubushakashatsi burakomeza mugushakisha ibimenyetso byafashwe mumutwe mugihe cyo gufata umwanzuro. Ibanga ritandukanye rishobora kuba ingirakamaro kuri domaine zoroshye nko gufata amashusho yubuvuzi, nubwo ibicuruzwa byiza bisigaye. Ishusho yemewe ntabwo ihagaze neza: inkiko zo mubihugu bitandukanye zirimo gusuzuma niba amahugurwa nibisohoka bibangamira uburenganzira, kandi ibisubizo birashobora gutandukana. Abashinzwe iterambere barashobora guhura nigitutu cyinshi kugirango bandike amakuru yamakuru yatanzwe kandi berekane uburyo bapima kandi bagabanya kugarura.
Ibaruwa y'amakuru y'ubuntu
Inkuru eshatu zizewe za AI buri gitondo cy'iminsi y'akazi, zanditse mu rurimi rworoshye. Ubuntu iteka ryose, nta kwamamaza.
Imeri imwe buri munsi y'akazi. Kwiyimura mu buryo bumwe. Ntitwigera tugurisha cyangwa ngo dusangire aderesi yawe.
Isuzume
Ibisubizo ako kanya kuri buri kibazo, n'icyemezo gishobora gusangirwa gifite ID yagenzurwa umaze gutsinda isomo.
Fasha uburezi bwa AI bw'ubuntu. AI Understanding ni umuryango utegamiye kuri leta wa 501(c)(3) — nta kwamamaza, nta kwishyura, iteka ryose. Tanga impano
Moderi yerekana amashusho ubusanzwe itanga amashusho mashya, ariko ubushakashatsi bwerekanye ko rimwe na rimwe bashobora kubyara amashusho yihariye yamahugurwa hafi ya yose, cyane cyane amashusho yagaragaye inshuro nyinshi mumibare yamahugurwa. Ibi byitwa gufata mu mutwe. Nibyingenzi kuburenganzira, kuberako hafi-kopi yumurimo urinzwe irashobora kuva mubyitegererezo, no kubanga, kuko ifoto yumuntu nyawe cyangwa scan yubuvuzi irashobora gusubirwamo.
Gufata mu mutwe bisobanura ibisubizo biri hafi-kopi yamashusho yihariye yo guhugura, kuruta amashusho mashya.
Gusubiramo kenshi kumashusho amwe bituma bikwiranye nurugero nyarwo inzira nziza yo kugabanya igihombo.
Bakuyemo kopi zirenga ijana hafi ya Stable Diffusion, harimo amafoto yabantu bamenyekana, basanga ibiciro biri hejuru muri Imagen.
Ishusho yafashwe mu mutwe igaragara buri gihe hatitawe ku mbuto, bityo rero uduce twinshi twa hafi-dusohoka ibendera ry'abakandida.
Amakopi yakuweho ni gake ugereranije nibisohoka byose, ariko abateye bashobora kwibanda kumashusho ashobora kuba yafashwe mumutwe.
Komeza wige
Abandi bayobozi batoranijwe kuriyi ngingo
HejuruUbuyobozi bukurikira
Kurwanya Diffusion Kurwanya hamwe na Turbo Model
Tekiniki