Midjourney banned all employees from rival AI firm Stability AI from its service indefinitely after detecting "botnet-like" activity suspected to be a Stability employee attempting to scrape prompt and image pairs in bulk, causing a 24-hour outage. This move comes after Midjourney faced criticism for using training data scraped off the Internet without permission. Stability AI CEO claimed the incident was unintentional and stated that his company doesn't need Midjourney's data, emphasizing their use of synthetic and other data.
Researchers at the University of Chicago have developed a data poisoning technique called "Nightshade" to disrupt the training process for AI models that scrape art without consent. The open-source tool alters images in ways invisible to the human eye, corrupting the training process and misidentifying objects within the images. The goal is to protect visual artists and publishers from having their work used without permission to train generative AI image synthesis models. The researchers hope that Nightshade will encourage AI training companies to license image datasets, respect crawler restrictions, and conform to opt-out requests.
Midjourney has unveiled version 5.2 of its AI-powered image synthesis model, which includes a new "zoom out" feature that allows maintaining a central synthesized image while automatically building out a larger scene around it, simulating zooming out with a camera lens. The update also includes an overhauled aesthetic system, a stronger "--stylize" command, a new "high variation mode," and a "/shorten" command. Midjourney's visual improvements have been called "stunning" and "mindblowing" by users, and the latest update is part of a series of quality improvements since March 2022.
Adobe has launched Firefly, an AI image synthesis generator that can create new images from text descriptions. Unlike other AI art models, Firefly has been trained solely on legal and ethical sources, making its output safe for commercial use. Adobe's commitment to ethical AI includes a "Do Not Train" tag for creators who do not want their content used in model training. Firefly's first model is trained on Adobe Stock images, openly licensed content, and public domain content where copyright has expired. Future models will leverage a variety of assets, technology, and training data from Adobe and others.
