Google demonstrated excerpts from its Gemini viral duck video, which is a rival to GPT-4, in a recent demo video. The fact that the video, “Hands-on with Gemini: Interacting with Multimodal AI,” had undergone substantial editing—a fact made clear in the video description—sparked controversy.
Must Read: Google Messages Update: Ultra HDR Image Support Unleashes a New Era in Messaging
The video, which was meant to showcase Gemini’s capabilities, caused a stir when Google acknowledged that, instead of responding in real time to drawings or changes in objects, it was simulating interactions using written prompts and altered still images. In response to the issue, Google DeepMind’s Vice President of Research and Deep Learning Lead, Oriol Vinyals, said, “I’m really happy to see the interest in our ‘Hands-on with Gemini’ video.” Vinyals explained that the film was meant to serve as inspiration for developers and to show possible multimodal user experiences with Gemini.
Gemini’s Development Process:
Vinyals clarified that in order for Gemini to forecast future behaviors, sequences of several modalities, including text and graphics, were provided. On December 13, developers will have access to Gemini Pro and can try out comparable strategies. For example, the knitting demonstration showed how to utilize Ultra with Gemini. Even with Vinyals’ promises, doubts remain over the demo’s legitimacy. An evolving drawing of a duck was shown in the original viral video, which showed Gemini’s reaction to several situations. It was eventually discovered, though, that neither voice activation nor real-time recording were used for this video.
Parmy Olson drew attention to the discrepancy between Google’s representation of Gemini’s capabilities and the actuality of its editing methods in a Bloomberg opinion post. Olson pointed out that the demo’s presentation gave the impression that Gemini could have natural voice conversations and react in real time to its environment.
The controversies surrounding Google’s Gemini popular duck video highlight how crucial transparency is when demonstrating the potential of artificial intelligence. Developers are excited to see Gemini Pro open to the public on December 13 in the hopes of seeing its full potential outside of staged demos, even while the argument rages on.
[Source]
