Google’s upcoming AI model, Gemini, is set to have multimodal capabilities, enabling it to process diverse types of data, including text, images, and graphs. Initially expected to be unveiled as a competitor to GPT-4, recent reports suggest that the launch date might be postponed to January 2024.
The Information, citing sources, indicates that Gemini encountered challenges with non-English queries, prompting Google CEO Sundar Pichai to delay the release. The report emphasizes the importance of supporting various languages to compete with or surpass OpenAI’s GPT-4, noting that Google has made progress in meeting this standard in some areas.
In a recent statement, Pichai mentioned the company’s focus on releasing Gemini 1.0 promptly while finalizing the largest version of the next-generation AI model. During the Google I/O event in May, the tech giant introduced Gemini as its most powerful model, based on PaLM 2 and developed by the Brain Team and DeepMind. Described as ‘multimodal,’ Gemini is designed to process different types of content, such as images and text, incorporating memory and planning capabilities.
Gemini is expected to integrate various AI models, coming in various sizes and capabilities. Google plans to use Gemini internally across its products and make it available to developers and cloud customers. Pichai highlighted the efficiency of the next-generation AI model, emphasizing its tools and API integrations.
The AI model is anticipated to power Google’s existing AI-driven products, including Search, Google Assistant, and the popular chatbot Bard. Despite the potential delay in the release, Google remains committed to delivering a highly efficient and versatile AI model with Gemini, showcasing advancements in multimodal capabilities and language support.
