The Download: The lab fighting exploitative AI, and plant engineering

Again in 2022, the tech neighborhood was buzzing over image-generating AI fashions, equivalent to Midjourney, Steady Diffusion, and OpenAI’s DALL-E 2, which might comply with easy phrase prompts to depict fantasylands or whimsical chairs made from avocados. 

However artists noticed this technological marvel as a brand new type of theft. They felt the fashions have been successfully stealing and changing their work.

Ben Zhao, a pc safety researcher on the College of Chicago, was listening. He and his colleagues have constructed arguably essentially the most outstanding weapons in an artist’s arsenal towards nonconsensual AI scraping: two instruments known as Glaze and Nightshade that add barely perceptible perturbations to a picture’s pixels in order that machine-learning fashions can not learn them correctly.

However Zhao sees the instruments as a part of a battle to slowly tilt the steadiness of energy from giant companies again to particular person creators. Read the full story.

—Melissa Heikkilä

Have we entered the golden age of plant engineering?

Within the Nineteen Sixties, biologists’ selective breeding of crops helped spark a interval of transformative agricultural innovation often called the Inexperienced Revolution. By the Nineties, the yields of wheat and rice had doubled worldwide, staving off bouts of recurring famine. 

The Inexperienced Revolution was so profitable that dire predictions of worse famine to return—fueled by alarming inhabitants progress—not appeared seemingly. However it had its limits—solely a lot yield may very well be coaxed from crops utilizing typical breeding methods.