Google has unveiled its latest artificial intelligence (AI) model, named Gemini, which the company claims is its largest and most capable language system to date. Gemini comes with three different-sized categories: Gemini Ultra, Gemini Pro, and Gemini Nano, each with its own unique capabilities.
Gemini Ultra is touted as the most adept model, specializing in highly-complex tasks. Gemini Pro, on the other hand, is designed to scale across a wide range of tasks, making it incredibly versatile. Lastly, Gemini Nano is optimized for tasks on mobile devices, making it the most efficient model for on-the-go use.
According to Demis Hassabis, CEO and co-founder of Google DeepMind, Gemini is a multimodal language model, meaning it can seamlessly understand, operate across, and combine different types of information, including text, code, audio, image, and video. This advanced coding and reasoning ability sets Gemini apart from previous AI models.
Google has thoroughly tested Gemini models across various domains, including image, audio, video, and mathematical understanding. The results have been impressive, with Gemini Ultra outperforming human experts on 30 out of 32 commonly-used academic benchmarks in the field. Moreover, Gemini Ultra achieved a remarkable score of 90.0% on a massive multitask language understanding (MMLU) test, covering 57 subjects including math, physics, history, law, medicine, and ethics.
Gemini’s capabilities extend beyond language understanding. It can extract information from hundreds to thousands of documents with sophisticated reasoning, paving the way for new breakthroughs across fields like science and finance. Gemini can even understand, explain, and create advanced code in programming languages such as Python, Java, C++, and Go.
In terms of safety, Google assures users that Gemini has undergone the most comprehensive safety evaluations to date. The company has conducted extensive research into potential risk areas, including cyber-offense, persuasion, and autonomy. Additionally, Google’s best-in-class adversarial testing techniques have been applied to identify critical safety issues in advance.
The rollout of Gemini begins with a fine-tuned version of Gemini Pro, which will be available in over 170 countries and territories starting December 6. On December 13, Gemini Pro will launch via the Gemini API in Google AI Studio and Google Cloud Vertex AI. Furthermore, Gemini Nano is set to debut on the Pixel 8 Pro, while Gemini will be integrated into various Google products such as Search, Ads, Chrome, and Duet AI in the coming months. Finally, Gemini Ultra will be made available to developers and enterprise customers in early 2023.
Overall, Google’s introduction of Gemini represents a significant leap forward in AI capabilities. With its advanced language understanding, reasoning, and coding abilities, Gemini has the potential to revolutionize various industries and drive new advancements at digital speeds.