The rapid evolution of artificial intelligence video generation was starkly illustrated this month with the latest iteration of the “Will Smith eating spaghetti” test, demonstrating a significant leap in realism and sophistication, according to reports from multiple technology outlets.
What began in 2023 as a “monstrous, pixelated mess” – an early attempt to render a video of the actor enjoying a plate of pasta using the ModelScope AI – has, within three years, transformed into a convincingly cinematic scene, complete with dialogue and interaction with another individual. The latest versions, notably those produced using Bytedance’s Seedance 2.0 and Kuaishou Technology’s Kling 3.0, have largely “passed” the test, generating videos that, while still identifiable as AI-generated, are far removed from the crude outputs of the initial attempts.
The “Will Smith eating spaghetti” scenario has become a benchmark for generative AI video, a “Hello World” equivalent for the technology, according to Mashable. The initial video struggled with maintaining a consistent depiction of Smith’s face from frame to frame. Now, the technology can convincingly portray the actor in a realistic setting, engaging in conversation. A recent demonstration using Kling 3.0, highlighted by Reddit users, showed Smith at a dinner table discussing the capabilities of the AI itself, a clear indication of the technology’s promotional use.
The speed of improvement has been described as “mind-blowing” by industry observers, with the 18-month period between the earliest and most recent iterations marking a dramatic shift in capabilities. This rapid advancement raises questions about the future of AI-generated video and its potential applications, as well as concerns about the authenticity of visual content.
While the technology has progressed, not all reactions have been positive. Hayao Miyazaki, the co-founder of Studio Ghibli, expressed critical views on AI art nearly a decade ago, a sentiment that continues to resonate as the technology becomes more prevalent. The evolution from “awkward infancy” to “photorealistic renderings” has prompted debate about the ethical and artistic implications of AI-generated content.
The development of these AI video models continues, with ongoing improvements in realism, and functionality. Further advancements are expected, but the long-term impact of this technology remains to be seen.
