Google Unveils Next-Generation AI Model, Gemini 1.5
Last week, tech giant Google made several major announcements, including the rebranding of ‘Google Bard’ to ‘Gemini’, the launch of the Gemini Advanced app, and the introduction of a new premium AI plan. Building on these developments, Google has now unveiled its latest AI model, Gemini 1.5, which boasts significant upgrades compared to its predecessor.
Gemini 1.5 introduces a longer context window and improved understanding and performance. Google CEO Sundar Pichai stated that the first model of Gemini 1.5, called 1.5 Pro, achieves a comparable quality to Google’s most advanced large language model (LLM), 1.0 Ultra, while using less computing power, making it more efficient.
The new model is built on a revamped version of the Mixture-of-Experts (MoE) architecture, enabling the AI to learn and selectively activate the most relevant pathways in its neural network. This enhances efficiency and effectiveness in processing large amounts of information.
One of the key improvements in Gemini 1.5 Pro is its processing capacity. The model can now handle up to one million tokens in production, a significant jump from the original 32,000 tokens for Gemini 1.0. This expansion allows Gemini 1.5 Pro to gather more information, resulting in better-informed responses.
Gemini 1.5 Pro also excels in processing various types of information, such as video, audio, and complex codebases. It can efficiently process codebases with over 30,000 lines of code or more than 700,000 words, making it suitable for developers and enterprises working with large datasets.
Google conducted extensive evaluations of Gemini 1.5 Pro to ensure its safety and responsible deployment. The model outperformed Gemini 1.0 Pro in 87% of the benchmarks used for developing Google’s LLMs. It also demonstrated strong performance in the Needle In A Haystack (NIAH) evaluation and Machine Translation from One Book (MTOB) benchmarks.
In a move to further enhance accessibility, Google is offering a limited preview of Gemini 1.5 Pro with a one million token context window to developers and enterprise customers through AI Studio and Vertex AI at no cost. Pricing tiers for wider release are expected to be introduced in the future.
With the introduction of Gemini 1.5, Google continues to push the boundaries of AI technology, providing users with more powerful and efficient tools for processing and generating information.
