Google’s new anything-to-anything AI model is wild
SMRTR summary
A stuffed deer named Buddy has become an unlikely test subject for Google's latest AI video tool. A journalist deepfaked the plush toy, and then herself, using Omni Flash, the newest model in Google's Gemini family, now available through its video platform, Flow.
The results are unsettling in their inconsistency and their convincingness. Buddy sprouted antlers he doesn't have. A honey jar shapeshifted throughout a single clip. And yet, a deepfake video of the journalist eating pasta fooled her own husband, a man who has seen her face nearly every day for a decade.
The technology isn't cheap. A $20 monthly plan burns through credits fast, and getting close to your vision often means costly rounds of back-and-forth with the model.
What lingers is a quiet dread. "I'm not sure anyone else would know the difference," she writes, "and that makes me feel weird." Shocking, yes. But the shock, she admits, is wearing off. And that might be the most unsettling part of all.
SMRTR provides this summary for quick context. The original article belongs to The Verge.
Read the original article