Google Unveils Gemini

December 7, 2023

After much anticipation and a bit of delay, Google has officially launched its latest AI system, named Gemini. Google CEO Sundar Pichai described it as the “biggest science and engineering effort” the company has undertaken, marking eight years of dedicated AI work.

Three Modes of Gemini

Gemini AI introduces three distinct modes: Ultra, Pro, and Nano. In the Ultra mode, the system employs the largest LLM (large language model) for intensive AI tasks. The Pro mode utilizes a smaller LLM, while the Nano mode uses the smallest LLM, suggesting potential local availability for computers and phones.

Competing with ChatGPT and AI Advancements

Google aims to compete, and possibly surpass, other AI systems, particularly OpenAI’s ChatGPT. Despite Google’s AI pioneer status through its investment in DeepMind, the company has faced challenges in keeping up with the rapidly advancing AI field, with ChatGPT setting new standards.

Pichai emphasized the significance of the AI transition, stating, “I believe the transition we are seeing right now with AI will be the most profound in our lifetimes.” Gemini represents a major leap forward for Google, showcasing state-of-the-art performance across leading benchmarks.

Gemini’s Human-like Performance

Gemini, especially the Ultra mode, outperforms humans in various language tasks, scoring 90.0% in massive multitask language understanding (MMLU). This achievement positions Gemini as the first model to surpass human experts in this domain, covering subjects like math, physics, history, law, medicine, and ethics.

Integration with Google Products and Services

As of Wednesday, Gemini is integrated into numerous Google products and services, including Bard, Google’s existing AI system. Bard will now incorporate a fine-tuned version of Gemini Pro for advanced reasoning, planning, and understanding. Additionally, Gemini Nano will power features like smart replies and the Recorder app on Google Pixel phones.

Multi-Modal AI Interaction

Gemini is touted as a multi-modal AI system, capable of interacting with users through various input methods, including text, images, and audio. Google plans to expand Gemini’s presence in more products and services like Search, Ads, Chrome, and Duet AI in the coming months.

« Previous Post Next Post »