Google has introduced ‘Gemini’ as its largest and most capable AI model. Built from the start as multimodal, Gemini can “seamlessly operate across and combine different types of information.” This includes text, image, audio, video, and code.
The first model, Gemini 1.0 comes in three ‘sizes’: the Gemini Nano aimed to run natively on-device to Android; a ‘fine-tuned version’ of Gemini Pro now empowers Google Bard; and Gemini Ultra capable for highly complex tasks that can understand and code in Python, Java, C++, and Go.
Google’s subsidiary, DeepMind with its CEO Demis Hassabis claims that Gemini outperforms OpenAI’s GPT-4 model in 30 out of 32 benchmarks.
Those benchmarks show narrow and even larger differences, with Gemini’s advantage of being multimodal by design.
Furthermore, Google specifically mentioned that the Pixel 8 Pro is the first smartphone engineered for Gemini Nano. It apparently powers new features like Summarize in the phone’s Recorder app and will roll out in Smart Reply to Gboard starting with WhatsApp.
“Gemini in Search” is also in development to make Search Generative Experience (SGE) faster with 40% reduced latency in English in the US.
Take a look at what Gemini can do with a video demo below:
Seeing some qs on what Gemini *is* (beyond the zodiac :). Best way to understand Gemini’s underlying amazing capabilities is to see them in action, take a look ⬇️ pic.twitter.com/OiCZSsOnCc
— Sundar Pichai (@sundarpichai) December 6, 2023
0 Comments
Leave a Reply