11/12/2023
Google unveiled its most advanced and versatile AI model, Gemini. However, it has come to light that the demonstration video showcasing its capabilities was, in fact, misleading. ?!
Google aimed to highlight the power of its new AI model, Gemini. The video depicted an impressive conversation between Gemini and a person, where Gemini seemed to respond seamlessly to combined speech and video. Complex image recognition and analysis were demonstrated by Gemini in a matter of seconds.
However, Google, in a surprising revelation on their own YouTube video, admitted that the conversation portrayed in the video did not occur as implied.
"For the purposes of this video, the latency has been reduced, and Gemini responses have been shortened," reads the disclaimer beneath the YouTube video.
Google reportedly confessed to Bloomberg that the interaction with Gemini did not take place as a spoken conversation but was instead assembled from a series of still images and video recordings. Text prompts were then generated to which Gemini provided responses.
"We created it to inspire developers," stated Orial Vinyals, research lead and head of deep learning at Google DeepMind, on X/Twitter.
On Google's developer blog, they posted the text prompts and still images that were used to guide Gemini, along with the responses it gave.
The actual dialogue between Google's developer and Gemini occurred through images and text.
During a closed press meeting before the official launch, Google presented three different models of Gemini: Nano, Pro, and Ultra. The most powerful version, Ultra, is currently not available. The Pro version of Gemini was introduced as an upgrade to Google's AI chatbot Bard, but it is only available in English and in 170 countries.
When comparing Gemini to OpenAI's ChatGPT during the press meeting, Google focused on three models of Gemini but avoided comparing the Pro version with OpenAI's latest model, ChatGPT 4.0. They only sought to draw comparisons with the earlier model, 3.5.
However, an examination of the comparison model released by Google reveals that the Pro version of Gemini performs worse in several aspects compared to ChatGPT 4.0, as opposed to ChatGPT 3.5. The Ultra version is purportedly stronger than the 4.0 model, but according to Bloomberg, the lead is very narrow, and OpenAI could potentially catch up with Google in the next update.
We follow.......
Gemini is our natively multimodal AI model capable of reasoning across text, images, audio, video and code. This video highlights some of our favorite intera...