Google introduces Gemini, its largest and most capable AI model

In a significant leap forward in the field of AI, Google DeepMind, a pioneer in AI research, proudly introduces Gemini, a revolutionary multimodal AI model designed to transform the landscape of AI applications. Gemini’s creation is poised to set new standards in the performance, flexibility, and safety of these apps.

Advertisement

Gemini, a collaborative effort across DeepMind and various other Google teams, is a state-of-the-art multimodal AI model capable of seamlessly understanding and processing diverse types of information, including text, code, audio, image, and video. Demis Hassabis, CEO and co-founder of Google DeepMind, described Gemini as “the most capable and general model we’ve ever built.”

Spearheading the unveiling, Google CEO Sundar Pichai expressed his enthusiasm for the transformative potential of Gemini. “The transition we are witnessing with AI will be the most profound in our lifetimes, far surpassing previous technological shifts,” he said. “Gemini has the power to create unprecedented opportunities for people worldwide, fostering innovation and economic progress on an unprecedented scale.”

People around the world also won’t need to wait before getting their hands on the amazing new technology, as Google is rolling out Gemini across its products starting with Bard and Pixel. In the coming months, Gemini will be available in more Google products like Search, Ads, Chrome and Duet AI. Meanwhile, developers can also start working with it via Google AI Studio or Google Cloud Vertex AI.

What Gemini is truly capable of

Gemini 1.0, optimized for three different sizes—Ultra, Pro, and Nano—has sophisticated reasoning capabilities that make it adept at extracting insights from vast amounts of data, leading to breakthroughs in fields ranging from science to finance. 

Notably, Gemini Ultra surpasses human expert performance on the Massive Multitask Language Understanding (MMLU) benchmark, demonstrating its unparalleled reasoning capabilities across various subjects. Its native multimodality sets it apart, allowing it to understand and reason about different inputs seamlessly. 

A more specific example of where Gemini excels would be coding tasks, understanding, explaining, and generating high-quality code in popular programming languages. Its prowess extends to competitive programming, where Gemini’s performance outshines its predecessor, AlphaCode, solving nearly twice as many problems.

Built for the future with responsibility and safety

Acknowledging the many different critical conversations surrounding the development of AI technology, Google reaffirms its commitment to responsible AI development with Gemini, incorporating the most comprehensive safety evaluations in the industry, including assessments for bias and toxicity. 

Gemini undergoes stress testing, including adversarial testing and evaluations for potential risks in areas like cyber-offense, persuasion, and autonomy. Responsibility with AI is also a collaborative effort, so Google is also partnering with the rest of the tech industry to define best practices and set the highest security and safety benchmarks.

As Google DeepMind continues to innovate from here, Gemini Ultra and future versions are on the horizon, promising even more advanced capabilities. The Gemini era signifies a transformative period for AI, with the potential to enhance creativity, extend knowledge, and transform the way billions of people around the world live and work every day.

For more information about this news, visit this blogpost.