Google has launched Gemini 2.0, a powerful upgrade to its AI model, which the company claims is twice as fast as its predecessor. This new version promises to revolutionize user experiences by assisting with searches, coding projects, and generating images and audio in multiple languages. According to Tulsee Doshi, Google’s Product Management Director, Gemini’s advanced abilities enable it to “think, remember, plan, and take action,” making it an indispensable tool for users.
As competition in the AI space intensifies, Google remains determined to maintain its dominance in search and advertising. OpenAI, a key competitor, has integrated search capabilities into ChatGPT, challenging Google’s supremacy. Both companies are racing toward developing artificial general intelligence—software capable of performing tasks on par with humans. Koray Kavukcuoglu, CTO of Google DeepMind, emphasized the importance of strategically selecting applications to maximize AI’s potential.
Google is also embedding Gemini 2.0 into its core products. Starting this week, the model will enhance AI-powered summaries in search results, delivering faster and more accurate answers to complex questions. Additionally, the company has made an experimental Gemini 2.0 Flash model available to developers. This model, designed for speed and efficiency, processes images and mimics human reasoning, pushing the boundaries of AI’s capabilities.
Among the new features is a "deep research" tool, enabling users to dive into topics with detailed reports. This AI-powered research assistant is available to Gemini Advanced subscribers. Gemini 2.0 Flash also introduces a chat-optimized version accessible globally, with plans to integrate the model into more Google products in 2024.
Google DeepMind, the company’s premier AI lab, has played a crucial role in these developments. It is expanding trials of Project Astra, an AI agent capable of processing visual inputs through a smartphone camera. During a demonstration, Astra successfully summarized text from books and provided insights into iconic art pieces, although it showed limitations, such as identifying books on a shelf. The agent’s conversational speed and accuracy have improved significantly, DeepMind researcher Greg Wayne confirmed.
In addition to Astra, Google is testing Mariner, a browser extension designed to assist with online shopping and digital organization. In a demo, Mariner added recipe items to a shopping cart, but Google is ensuring that users retain control over key decisions to prevent errors. “It’s not just about convenience but about trust,” said Helen King, Google DeepMind’s senior director of responsibility.
Google is also exploring other experimental AI agents. One, named Jules, assists engineers by fixing software bugs and automating programming tasks. Another unnamed agent for video games offers real-time guidance based on gameplay. Both are in the early testing stages, showcasing Gemini 2.0’s diverse applications.
Despite concerns about the financial returns on AI investments, Google remains optimistic. Kavukcuoglu highlighted Gemini 2.0’s rapid progress and efficiency compared to a year ago, calling it a significant leap forward.