Google Gemini: A New Era in Artificial Intelligence
Google has just announced the launch of Gemini, an innovative AI model considered the most capable of all. Integrated with the Bard chatbot, the new AI tool has been trained on a variety of media, including video, images, audio and text, marking a major advancement in AI technology and can be considered Google's most significant algorithm since PageRank.
What is Google Gemini?
It's the most recent Google innovation in the AI space. Built with the idea of overcoming existing models, such as Chat GPT, promises to be not only more advanced in terms of natural language processing capabilities, but also more efficient and adaptable to different contexts and applications.
Created by Google DeepMind, under the leadership of CEO and co-founder Demis Hassabis, Gemini is a demonstration of Google's ongoing commitment to prioritizing artificial intelligence in its operations.
This model displays a remarkable range of abilities, particularly excelling in multimodal understanding. This functionality enables the fluid processing and integration of various types of information, such as text, code, audio, images and videos.
Implementation and Availability
Gemini has already started to be integrated into Google Bard, chatbot similar to ChatGPT, for English-speaking users in more than 170 countries. Furthermore, will be available to developers via the Google Cloud API from December 13th. A compact version of the model is already being used to suggest replies to messages on Pixel 8 smartphones, with plans to expand it to other Google products, such as generative ai research, ads and Chrome.
Gemini Templates
Google introduced three versions of Gemini:
Gemini Ultra: The largest and most capable model.
GeminiPro: Medium size with average capabilities.
Gemini Nano: A smaller and more efficient version.
Google's Bard will now be powered by Gemini Pro and promises advanced reasoning and planning. A specialized version of Gemini Pro is also being transformed into a new version of AlphaCode, a Google DeepMind coding tool.
Superior performance
The Gemini Ultra, which will be released next year, achieved a score of 90% on the MMLU benchmark, beating other models including the GPT-4. This performance puts Gemini at the forefront across a wide range of benchmarks, setting new standards in machine learning research.
Responsibility in AI Development
Google has emphasized carrying out the most comprehensive security testing for Gemini, considering the model's broader capabilities. The project involved the collaboration of researchers and engineers from across Google, using an updated version of the company's custom silicon chips, Tensor Processing Units (TPUs).
Conclusion
The launch of Gemini is a significant moment in the development of AI, showcasing Google's dedication to innovating and leading in the field of artificial intelligence. With Gemini, the company not only responds to the success of ChatGPT, but also sets a new standard in terms of AI capability and versatility.