During the Second World War, Hollywood actress Hedy Lamarr made many breakthroughs in the field of technology, and her work has lead to modern innovations like Wi-Fi, CDMA, and Bluetooth technology.
Now, another actress has entered the scientific community and it might come as a bit of a surprise. Kristen Stewart, best known for Twilight and Snow White and the Huntsman, has co-authored a research paper, titled “Bringing Impressionism to Life with Neural Style Transfer in Come Swim.”
The paper focuses on a type of artificial intelligence called machine learning, which gives the impression of a shot on film resembling a painting in a certain style, similar to impressionism or pointillism. Stewart likely became involved with the project thanks to her directorial piece, “Come Swim,” which she intended to resemble a painting, and had earned her place as a co-author from her directorial work, which went into the project.
The lead author in the piece is Bhautik Joshi, with Stewart as the second, and David Shapiro is the third.
“The algorithm is essentially a black box,” Joshi said. “Randomly sampling all these looks — that would get us nowhere. So we wanted to approach it in a structured way. We said, ‘What is a reasonable range for this?’ until we converged on the look, and made our iterations more predictable…You basically start with something — present [the] director with a starting point, and you iterate to get the imagery to a point to get the emotional response you want.”
Eventually, the computer generated an image Stewart was happy with — one of a man lying on his back in the water.” The technology involved was theorized sometime in the 1940s, but developments in the field took time and means to test the hypothesis put forward were not yet available. What’s more, the process is said to be time-consuming.