Google is already preparing to launch a new AI system Gemini

The launch of ChatGPT late last year prompted many tech giants, such as Google, to develop their own AI chatbots like Bard. However, according to a new report from The Information, it looks like Google is already preparing to launch Bard’s successor named Gemini.

Developed in collaboration with Google and DeepMind, the new chatbot could reportedly launch as early as this Autumn and potentially reshape the AI chatbot landscape. This is because an insider source revealed that Google’s Gemini aims to merge the textual capabilities of large language models (LLMs) with the creative potential of AI-generated images. This innovative approach has the potential to unlock a world of new possibilities, as Gemini will not only function as a text generator but also as a contextual image creator. Furthermore, Google’s vision also includes features such as chart analysis, the creation of graphics accompanied by textual descriptions, and the ability to manipulate software using text or voice commands.

Although the specific details about how Google plans to implement Gemini in its services remain unclear, the company could draw inspiration from Microsoft’s strategy by integrating Gemini into Google Docs and Sheets as a helping hand to users who lack technical expertise.

Moreover, to help make Gemini the next leap forward in the world of generative AI, Google has assembled a team of prominent figures from Google Brain and DeepMind. Among them are Paul Barham, a senior Google researcher, visionary Tom Hennigan from DeepMind and Google’s co-founder, Sergey Brin, who plays a pivotal role in the evaluation and refinement of the AI models.

How is Google training Gemini?

Instead of using traditional methods of training AI models, which often involve collecting data from social media platforms, Google has chosen a unique path by using YouTube video transcripts to train its Gemini AI model. This approach taps into the extensive content library of YouTube, enabling Google’s model to amass knowledge across an extensive array of domains.

Nevertheless, the path toward Gemini will not be without challenges due to Google’s recent legal struggles related to the use of copyrighted data for AI model training. However, to address this, the company is committed to maintaining compliance with copyright regulations and has also proposed a new “opt-out” feature, providing publishers with the ability to prevent their works from being utilized for AI training purposes.