Gemini, a next-generation large-scale language model (LLM) being developed by Google, has begun to compete for the lead in generative AI by creating a major shift in perception that will herald the era of artificial intelligence (AI) in earnest. Google recently announced its latest AI model, Gemini, a multimodal model capable of processing and generating multiple forms of information, including text, images, audio, video, and code. Gemini comes in three variants: Ultra, Pro and Nano, each tailored to specific capabilities and applications. Google CEO Sundar Pichai announced Gemini at the Google AI Summit on December 6, where he described it as "the most powerful and general model we've ever built.
Gemini is the result of a collaboration between Google's AI teams, DeepMind and Google Brain. It builds on their previous work on PaLM 2, a language model that powers many of Google's products and services. The model is designed to be highly efficient for tool and API integration, and is built to enable future innovations such as memory and planning. It is expected to be a game-changer for research and innovation, as well as education and entertainment.
Gemini Ultra will be released after completing the testing phase. It is the largest and most powerful version, with over 1.5 trillion parameters and cutting-edge performance on many leading benchmarks. Gemini Ultra can perform complex tasks that require reasoning, understanding, and creativity, such as answering questions, writing essays, generating code, and creating images from text. Gemini Ultra can also learn from and transfer knowledge across multiple modalities and domains. For example, it can generate a realistic image of a person based on a textual description, or write a summary of a video clip.
Gemini Pro is the most flexible and accessible version, designed for a variety of tasks that enhance human communication and creativity. Gemini Pro can help users write, rewrite, and generate creative content such as stories, poems, songs, jokes, and more.
In addition, Gemini Nano is a model designed to run on the Pixel 8 Pro, which Google unveiled in October this year, and is designed to summarize, translate, and parse text on smartphones, chat apps, and perform tasks that require AI processing without connecting to external servers to find relevant information and sources.
Controversy over the Gemini disclosure
Despite Google's passion for Gemini, its release is not without controversy. Critics pointed out that the demonstration video showing Gemini's capabilities was pre-edited and that the video was not live. The Google video controversy can be said to have sparked fierce competition from OpenAI's ChatGPT and the recently released OpenAI-MS, Meta AI, and others. The criticism of Gemini's video also highlights the importance of transparency in AI capabilities, and shows that users will approach the technology with an awareness of its true capabilities and limitations.
Gemini is an important step forward in generative AI, a technology that has the potential to revolutionize the way we interact with technology. Experts argue that the Gemini debate is an opportunity to advance development by overcoming controversy, as Google has differentiated itself by pouring enormous investment costs.
Gemini's multimodal capabilities and Google's vast exclusive learning data from various services aim to challenge open AI's GPT-4 advantage in the field of Bard generative AI equipped with Gemini. However, Gemini's ability to do too many things does not mean that the actual capabilities of the technology are limitless. It is important to broaden the awareness of the real capabilities and limitations of the technology and to approach it.
"Gemini is not magic, but it is magical," said Pichai, CEO of Google, emphasizing the use of responsible technology.
