philip lelyveld The world of entertainment technology

Kaleidoscope Vision: Seeing the Multifaceted Future of AI

...

The Caltech team designed an experimental method specifically to avoid these issues. Rather than testing algorithms using images of real people collected from random sources, the researchers used AI to generate a dataset of realistic human face images that were systematically varied across age, gender, race, facial expression, lighting, and pose. They also created a dataset of text prompts that described social perception based on findings from psychological research (e.g., “a photo of a friendly person,” and “a photo of dishonest person.”)

...

Alvarez also serves as co-director of Caltech’s Linde Center for Science, Society, and Policy (LCSSP) along with Professor of Philosophy Frederick Eberhardt. One of the center’s functions is to connect efforts across the Institute that aim to understand and steer the responsible implementation of AI. The LCSSP also provides scientific expertise to inform policy on pressing societal issues such as the implications of biotechnology as well as climate change and sustainability.

...

“We’re at this tipping point,” Alvarez adds. “If attitudes become polarized along partisan lines, it makes it very, very difficult for policymakers to effectively deal with AI.” Eberhardt says the LCSSP aims to build a bridge between Caltech researchers and policymakers “that will ensure a more secure integration of these two communities.”

...

“I am a Russianist, and I often raise with students this question about the development of AI technologies that enable autocratic regimes to track and persecute political dissidents,” Dennison says. “I point out to them how important it is to acknowledge the dark side of this advancement and encourage them to be clear about the larger implications of what they want to work on. It’s fine to argue that the positives outweigh the negatives. But, as with nuclear technology in the 20th century, there are important debates around these questions. It can be an uncomfortable conversation, but it is necessary.”

...

While the ethical debates, regulatory landscapes, and shifting social realities of AI may be complex, Perona says Caltech students and scientists are well equipped to work through them together while also continuing to tackle the hardest scientific questions. “There are questions that the AI industry is not interested in because there is no market,” he says. “We can work on them here at Caltech. In fact, this is probably the best place on Earth to do it.”

See the full story here: https://magazine.caltech.edu/post/kaleidoscope-vision?utm_medium=email&utm_campaign=magazine-spring25&utm_source=weekly-newsletter&utm_content=&utm_term=