Tech boss Google has unveiled Gemini, its formidable contender against OpenAI’s GPT-4. The language model was first teased at the I/O developer conference in June and the model is now launching to the public. The makers of Gemini describe it as said to be a “significant leap forward in how AI can help improve our daily lives.”
Google calls Gemini its most powerful and versatile AI model to date. It is a multimodal AI model, designed to seamlessly understand and operate across various types of information, including text, code, audio, image, and video.
“These are the first models of the Gemini era and the first realization of the vision we had when we formed Google DeepMind earlier this year. This new era of models represents one of the biggest science and engineering efforts we’ve undertaken as a company. I’m genuinely excited for what’s ahead, and for the opportunities Gemini will unlock for people everywhere,” Google CEO Sundar Pichai said during the announcement.
ALSO READ l Meta to extend verified tag to businesses in India
Gemini comes in three versions. First, there’s Gemini Nano, a lighter version made to run directly on Android devices without needing an internet connection. Then there’s Gemini Pro, a more robust model that’s set to drive many Google AI services and serves as the core for Bard. Lastly, there’s Gemini Ultra, the most potent one, designed mainly for demanding tasks in data centers and business applications.
Google says Gemini is natively multimodal, trained from the ground up to understand different modalities. This enables the model to excel in tasks that require conceptual and complex reasoning, setting it apart from existing models. It is trained to recognize and understand text, images, audio, and more simultaneously, making it adept at explaining complex subjects like math and physics.
OpenAI’s GPT-4 Vs Google Gemini
In a head-to-head comparison between OpenAI’s GPT-4 and Google’s Gemini, Google claims superiority based on a thorough analysis of 32 benchmarks. According to DeepMind CEO Demis Hassabis, Gemini outperforms GPT-4 in 30 out of these 32 benchmarks.
Availability and future plans
Gemini 1.0 is rolling out across Google products, including Bard, Pixel 8 Pro, and Search. Developers and enterprise customers can access Gemini Pro via the Gemini API in Google AI Studio or Google Cloud Vertex AI starting from December 13. Gemini Ultra is currently undergoing trust and safety checks and will be available for early experimentation in the coming months.
Follow FE Tech Bytes on Twitter, Instagram, LinkedIn, Facebook.