[ad_1]
Google has introduced its ‘largest’ and ‘most capable’ large language model, which goes by the name of Gemini, to the world. Dubbed Gemini 1.0, the model is the “start of a new era” and is the first of “new generations of AI models,” reads the blog, penned by Google CEO Sundar Pichai and Demis Hassabis, CEO of Google’s AI unit DeepMind.
Gemini is a combined multimodal AI modelconsisting of fundamental models.It is being pre-trained and fine-tuned on various modalities, allowing it to understand and reason about different inputs seamlessly and perform better than existing multimodal models. The system is proficient in programming, and it is twice as fast as its predecessor, putting its performance above 85% of the previous competition’s participants.
Ultra, Pro, Nano – three different models for different purposes
Gemini is not a single language model but three models – Gemini Nano, Gemini Pro, and Gemini Ultra. Each has its own use case.
Gemini Ultra is the largest and most capable one, designed for highly complex tasks, and is ideal for use in data centers and enterprise applications. Meanwhile, Gemini Pro is not as large as the Ultra but is a beefier one that can scale across a wide range of tasks and will be used to power many Google AI services. Then, there is Gemini Nano, which is a rather lighter model that can be run natively and offline on Android devices. It is also the most efficient model and is perfect for specific tasks and mobile devices.
Bard gets a ‘Pro’ upgrade
Bard, which is powered by Google’s LaMDA model, is being upgraded to Gemini Pro. Starting today, users can experience the advanced reasoning, planning, and understanding capabilities of the Gemini-powered Bard. The new and improved Bard will be available in the same 170 countries as this version. An even more advanced version called “Bard Advanced” is in the works for next year, which will be powered by Google’s largest and most advanced model, Gemini Ultra.
Developers and enterprise customers can try out the Gemini Pro through Google Generative AI Studio or Vertex AI in Google Cloud starting December 13.
Gemini is a combined multimodal AI modelconsisting of fundamental models.It is being pre-trained and fine-tuned on various modalities, allowing it to understand and reason about different inputs seamlessly and perform better than existing multimodal models. The system is proficient in programming, and it is twice as fast as its predecessor, putting its performance above 85% of the previous competition’s participants.
Ultra, Pro, Nano – three different models for different purposes
Gemini is not a single language model but three models – Gemini Nano, Gemini Pro, and Gemini Ultra. Each has its own use case.
Gemini Ultra is the largest and most capable one, designed for highly complex tasks, and is ideal for use in data centers and enterprise applications. Meanwhile, Gemini Pro is not as large as the Ultra but is a beefier one that can scale across a wide range of tasks and will be used to power many Google AI services. Then, there is Gemini Nano, which is a rather lighter model that can be run natively and offline on Android devices. It is also the most efficient model and is perfect for specific tasks and mobile devices.
Bard gets a ‘Pro’ upgrade
Bard, which is powered by Google’s LaMDA model, is being upgraded to Gemini Pro. Starting today, users can experience the advanced reasoning, planning, and understanding capabilities of the Gemini-powered Bard. The new and improved Bard will be available in the same 170 countries as this version. An even more advanced version called “Bard Advanced” is in the works for next year, which will be powered by Google’s largest and most advanced model, Gemini Ultra.
Developers and enterprise customers can try out the Gemini Pro through Google Generative AI Studio or Vertex AI in Google Cloud starting December 13.
Gemini is coming to Pixel phones, too
In the meantime, Pixel 8 Pro, Google’s newest phone, is getting some new AI features powered by Gemini Nano. The first is an auto-summarisation feature in the Recorder app, and the second is Smart Reply. These features work offline, but the Smart Reply is currently only available in English and can be used with WhatsApp. In terms of what’s next, a Gemini-powered Bard is planned for Google Assistant next year.
As for those who want to give the Ultra a try, it will not be made publicly available until at least 2024, as it needs some additional red-team testing.
According to Pichai, Gemini will eventually be integrated into Google Search, Chrome, and more.
[ad_2]
Source link