Artificial Intelligence
Artificial Intelligence
Out of context: Reply #1237
- Started
- Last post
- 1,324 Responses
- wagshaft0
Do of ya'll who work at agencies have an AI policy? Looking to start drafting one for the place I work.
- you're gonna need a lawyer...jonny_quest_lives
- seriously.jonny_quest_lives
- currently we can't even use Adobe generative products without client authorization.jonny_quest_lives
- It's more an ethic stance. A point of view around AI. Similar to how agencies have a DEI stance.wagshaft
- we've tested midjourney but we can't upload any client IP into it becuase of their lack of transparency as to how much they retain or reweight their modelsjonny_quest_lives
- I've seen client's asking for our POV on AI during pitches and I'm seeing a need to make sure we have a good answer.wagshaft
- all boils down to client authorization i.e.- if they are willing to endure the public fallout if goes sidewaysjonny_quest_lives
- agency management seems to feel that an honest response is best that at the end of the day it boils down to the strength of the concept not the execution.jonny_quest_lives
- whatever that means but the legalities and public blowback are real and can be brand damaging.jonny_quest_lives
- some clients are already more prepared.they tend to have their own legal departments that have already updated NDA's and data safety termsjonny_quest_lives
- we had a flare up a month ago with shutterstock over this mess: https://petapixel.co…jonny_quest_lives
- Art directors were sorting by "fresh content" and apparently all this abuse imagery started hitting the feed.jonny_quest_lives
- not a client but we have received similar guidance on projects: https://i.imgur.com/…jonny_quest_lives
- ^ jeezos that's grim.PhanLo
- Surely they would be able to somehow trace the person who uploaded that stuff?PhanLo
- the user used shutterstock's Ai image generator to generate the images. the user who exploited the tool was banned apparentlyjonny_quest_lives
- larger ethical issues are still in play though. how was shutterstock's Ai image generator trained? apparently LAION-5B contains CSAM so any model trained onjonny_quest_lives
- LAION-5B is capable of generating offensive content... fruit of a poisonous tree comes to mind.jonny_quest_lives
- https://cyber.fsi.st…jonny_quest_lives
- https://www.theregis…jonny_quest_lives
- we just requested our first Chat GPT Plus accounts at my work , it went straight to our cyber security team to vet it , not sure why but thats what im seeingBattleAxe