Google has introduced Gemini, its latest and most capable generative AI model, available in three iterations: Ultra, Pro, and Nano. Gemini 1.0 is optimized for different sizes and is the result of large-scale collaborative efforts by teams across Google, including those at Google Research and Google DeepMind. It is a multimodal model capable of seamlessly understanding and combining various types of information, such as text, code, audio, image, and video.
The three versions of Gemini—Ultra, Pro, and Nano—serve different purposes. Gemini Ultra, the largest and most capable model, outperforms human experts on massive multitask language understanding (MMLU). It achieves a score of 90%, exceeding state-of-the-art results on 30 of the 32 widely-used academic benchmarks in large language model research.
Gemini Pro is designed for scaling across a wide range of tasks, while Gemini Nano is tailored for on-device tasks. The sophisticated multimodal reasoning capabilities of Gemini 1.0 enable it to understand, explain, and generate high-quality code in popular programming languages like Python, Java, C++, and Go. Additionally, Gemini can be utilized as the engine for advanced coding systems.
Google has trained Gemini 1.0 at scale using its AI-optimized infrastructure and Tensor Processing Units (TPUs) v4 and v5e. The introduction of Gemini represents a significant advancement in generative AI, showcasing state-of-the-art performance across various benchmarks and applications, from natural language understanding to complex coding tasks.
(With Agency Inputs)
Read also| Google Settles for $27 Million with Employees Over Unfair Labor Practices
Read also| Survey Finds 92% of Indian Firms See GenAI Tools as Potential Security Risk
