AI poison - a way for artists to fight back, with what is essentially malware embedded in their images so that when an AI algorithm scrapes artists’ pictures (or words!) the machine “learns” nonsense…
Todays signal looks at radical adversarial alternatives to flawed copyright laws as a way for creators to take on machines.
“Poisoned data samples can manipulate models into learning, for example, that images of hats are cakes, and images of handbags are toasters. The poisoned data is very difficult to remove, as it requires tech companies to painstakingly find and delete each corrupted sample. “
Your thoughts on the ethics and efficacy of copyright in the face of planetary scale technology? And the ethics and efficacy of deliberately naughty data?
Can deliberately bad data protect us? Or is it a threat that can be used by bad actors to get generative AI platforms to push our not just funny-bad, but downright dangerous and inaccurate outputs?
Do you enjoy seeing humans battle machines? Or should we be trying harder to make friends with our tools? (I love a bit of anarchy and chaos but the stakes here are growing higher)
Is art going to go more analogue - are artists going to go AWOL offline and aim to become non machine readable as these battles become more pitched? Is AI art going to be worse in the future? Or will the algorithms learn around the Nightshade?
Read more:
Poison Apple Pencil : https://www-technologyreview-com.cdn.ampproject.org/c/s/www.technologyreview.com/2023/10/23/1082189/data-poisoning-artists-fight-generative-ai/amp/
Share this post