- By Vikas Yadav
- Sat, 09 Dec 2023 06:02 PM (IST)
- Source:JND
Gemini AI Demo Video: Soon after Google announced its ace AI model Gemini, a video demo has recently become the centrepiece of attention across media reports. First highlighted in a Bloomberg op-ed, the hands-on video is currently the hot topic of the internet. The first thing noted by Columnist Parmy Olson is that the model would have probably taken longer than portrayed in the video.
Secondly, Olson noted that the demo was not filmed in real-time with voice, unlike what is showcased in the clip. This video was made "using still image frames from the footage, and prompting via text," citing a Google spokesperson, the opinion piece read. Simply put, the voice in the demo was reading out prompts but these were not fed to the model in real time.
"For the purposes of this demo, latency has been reduced and Gemini outputs have been shortened for brevity," Google stated in the video description. Plus, the output from Gemini presented in the video was recorded from still images and not changing images, as showcased in the clip. This is contrary to what Google's demo indicates.
The controversy around the twisted video of Gemini's demo is also a reminder of the tech giant's shaky track record with demos. (Image:Google)
Viewers can infer that the two-way conversation happened in real time with voice-based and dynamic imagery as live inputs. However, this stands wrong based on the tech giant's remarks to the news outlet. Plus the author noted that the model used was probably Gemini Ultra, the top-of-the-line iteration which has not been released yet.
Soon after the reports pointed this out, taking to X (formerly Twitter), Oriol Vinyals, VP of Research and Deep Learning Lead at Google Deepmind, said: "We gave Gemini sequences of different modalities — image and text in this case — and had it respond by predicting what might come next."
"The knitting demo used [Gemini] Ultra. All the user prompts and outputs in the video are real, shortened for brevity. The video illustrates what the multimodal user experiences built with Gemini could look like. We made it to inspire developers," Vinyals added. The controversy around the twisted video is also a reminder of the tech giant's shaky track record with demos. The last viral instance was with Duplex.