Artificial Intelligence and Images

The Glaze Project is a system that disrupts style mimicry. According to the developers at the University of Chicago, Glaze minimally changes the underlying machine learning algorithms. This article explains the process: What is Glaze?

AI Poisoning Tools are being developed that will help artists protect their digital art works. 

• Nightshade, the free tool that ‘poisons’ AI models, is now available for artists to use
  • Nightshade, by computer scientists on the Glaze Project at the University of Chicago, identifies what is in a given image then alters the image at the pixel level by applies a tag. This results in other AI programs seeing something other than what is actually there. 
• AI poisoning could turn open models into destructive “sleeper agents,” says Anthropic
  • Research is being done on AI large language models (LLMs) that can be deceptively used to output a vulnerable code. Anthropic's research paper is here.
• Images altered to trick machine vision can influence humans too
  • Adversarial images are ones that are altered by a procedure that causes an AI model to misclassify the image contents. Behavioral experiments showed that human brains also pick up on weak perceptual signals when presented with choices.