Google Bard, its AI chatbot, has been receiving regular feature updates from Google. It hasn’t, however, proven to be a formidable rival to ChatGPT and other AI chatbots that have GPT-3 and GPT-4 installed. Google has now disclosed that the Bard is now powered by the Gemini Pro, a completely new AI model from Google.
As per Google, the Gemini is the most proficient AI model the company has ever developed. This versatile AI model can process and integrate text, code, audio, images, and videos with ease. It is designed to run on a variety of platforms, including mobile devices and data centers.
Gemini Ultra, Pro, and Nano
Due to its multimodal capabilities, Google Gemini 1.0 can comprehend complex information and provide relevant answers. It is available in three different sizes: Gemini Ultra for demanding tasks, Gemini Pro for versatile usage, and Gemini Nano for streamlined on-device operations.
Updated with Gemini Pro, the Google Bard can now perform more sophisticated planning, reasoning, understanding, and other tasks. It will be available in English in over 170 countries and territories. The Pixel 8 Pro will soon receive an update to run Gemini Nano, which will bring features like Summarize in the Recorder app, Smart Reply in Gboard, and WhatsApp. This is the biggest update to date for the Google Bard.
In terms of Gemini Ultra, it can handle a lot of different tasks: natural image, audio, and video understanding; mathematical reasoning; multimodal problem-solving; among other things. Google reports that Ultra has outperformed human experts on MMLU, a benchmark that evaluates knowledge and problem-solving skills, and that it scored exceptionally well on MMMU, a multimodal reasoning benchmark. Moreover, Ultra has outperformed previous models on image benchmarks without the need for object character recognition, demonstrating its native multimodality and complex reasoning abilities.
Before a wider rollout to developers and enterprise customers in early 2024, Google intends to make Gemini Ultra available to a small subset of users for early testing and feedback. Bard Advanced, a new AI experience powered by Gemini Ultra, will also launch in early 2024 to give users access to cutting-edge capabilities.
A multi-layered approach to safeguarding against harmful content is used by Gemini, which uses specialized safety classifiers to identify, label, and filter out harmful content, such as violence and negative stereotypes. Additionally, Google has introduced Cloud TPU v5p, the most powerful TPU system to date, enabling faster training of large-scale generative AI models and an acceleration of the development of new products and capabilities.