Google presents Gemini: a revolutionary AI model capable of surpassing ChatGPT-4
Google has unveiled a new advanced artificial intelligence (AI) model called Gemini, which, according to Google, can significantly outperform the current leader Chat GPT. The company is placing huge bets on the new model in 2024 and calls it a breakthrough in custom AI.
According to Sundar Pichai, Google CEO, it is time for a new era of AI in the company, which he calls the “Gemini era”. Gemini is Google’s latest development in the field of large-scale language models. The project was announced back in June at the I/O developer conference, and now it is available to the public. Pichai and Google DeepMind CEO Demis Hassabis describe Gemini as a significant advancement in the development of AI models that will affect almost all Google products. Pichai notes: “What’s special about this moment is that by working on one core technology and improving it, we can immediately impact all of our products.”
Gemini is not just one AI model. There is a lighter version called Gemini Nano, designed for offline use on Android devices. A more powerful version, Gemini Pro, will soon become the basis for many Google AI services, as well as for Bard. There is also the most powerful Gemini Ultra model designed for data centers and corporate use.
Google is currently implementing the model in several ways: Bard is now powered by Gemini Pro, and Pixel 8 Pro users will get new features with Gemini Nano. Gemini Ultra is scheduled for next year. Developers and enterprise customers will be able to access Gemini Pro via Google Generative AI Studio or Vertex AI in the Google Cloud starting December 13. Currently, Gemini is only available in English, but other languages are planned to be added. Pichai points out that the model will eventually be integrated into Google search, advertising products, the Chrome browser, and more.
OpenAI launched ChatGPT a year ago, and it quickly became an important part of the AI world. Now Google is preparing to meet the challenge.
It’s important to emphasize that Gemini excels at interacting with video and audio, thanks to its multimodality. Google didn’t create separate models for images and voice, like OpenAI did with DALL-E and Whisper, but developed a universal multi-sensory model from the start. Hassabis says that this is part of a broader development plan for Gemini, which will eventually gain more capabilities and become more accurate and realistic.
Gemini is not only more powerful, but also more efficient than previous models. It has been trained on Google’s Tensor Processing Units, making it faster and cheaper to use. Google is also launching a new version of its TPU system, TPU v5p, designed for training and use of large-scale models in data centers.
Pichai and Hassabis see the launch of Gemini as the beginning of a larger project and a game-changing move. Gemini may become not only Google’s answer to ChatGPT, but also the beginning of a new era in artificial intelligence technologies.
