OpenAI has been revealing daily announcements since December 5, 2024, including the launch of the o1 model, ChatGPT Pro subscription, and Sora video generator, with more updates like GPT-4.5 expected. In response, Google introduced Gemini 2.0, a multimodal language model that enhances voice assistant capabilities and aims to create AI agents for human-like conversations. Additionally, Google announced Projects Astra and Mariner, focusing on advanced AI interactions and efficient web browsing.
OpenAI’s Advent Calendar of Announcements
Since December 5, 2024, OpenAI has been unveiling exciting news every day. The mastermind behind ChatGPT, known for his flair for the dramatic, has introduced what he dubs an advent calendar to keep the conversation flowing. So far, he has rolled out the innovative o1 model, the ChatGPT Pro subscription, and Sora, a cutting-edge video generator. Anticipation builds as more updates, including the GPT-4.5 model, are on the horizon.
Google’s Response: The Launch of Gemini 2.0
Amidst OpenAI’s buzz, competitors are stepping up their game to share the limelight. Following xAI’s recent launch of a new image generator, Google has now made its move with several notable announcements. On December 11, the tech giant introduced Gemini 2.0, touted as its most advanced language model yet.
Gemini 2.0 is a sophisticated multimodal language model capable of generating both images and sound, enhancing its potential as a more advanced voice assistant. Google envisions innovative applications, including the creation of “AI agents” that can engage in human-like conversations.
While the most advanced iterations of Gemini 2.0 are currently in testing and not yet available to the public, the Gemini 2.0 Flash—designed for quick responses—is now free to all users of Google’s chatbot. This new model will soon be integrated into AI Overviews, enriching Google searches with AI-generated summaries. This move aims to reclaim ground from competitors like Perplexity and ChatGPT Search.
According to Google, the Gemini 2.0 Flash version outperforms Gemini 1.5 Pro despite being engineered for simpler tasks. Initial benchmarks are promising, indicating that the forthcoming Gemini 2.0 Pro and Ultra versions could rival leading models in the market. The capability for native audio generation opens fresh avenues for Google’s assistant, making it potentially more “emotional,” thus aligning better with ChatGPT Voice.
Exciting Developments: Project Astra and Project Mariner
In addition to Gemini 2.0, Google is sharing insights into its future endeavors. One of the highlights is Project Astra, an AI that can “see” and vocally comment on real-time screen activity. With the enhancements from Gemini 2.0, Astra can engage in natural conversations and seamlessly switch between languages. It can also utilize external tools such as Search, Lens, and Maps, thanks to its integrated platform capabilities. Furthermore, Astra boasts a quick response time and can retain memory for up to 10 minutes per conversation.
Looking ahead, Astra is expected to become the default voice assistant within the Gemini application. Google has also hinted at testing smart glasses outfitted with this voice assistant, confirming speculations about such a product.
Another significant announcement is Project Mariner, a Chrome extension designed as a sidebar that facilitates web browsing on your behalf. Users simply issue a command, and Mariner swiftly navigates the internet to extract relevant information. Its speed and efficiency in filling out forms, along with the option to pause at any time, set it apart from traditional browsing.
Additionally, Google has shared intriguing ideas for utilizing Gemini 2.0, including “Jules,” an agent focused on assisting developers, and the concept of creating agents that can navigate open-world video games to aid their creators. These agents could perform web searches to locate tasks within a game map, and assist novice players by automatically interacting with the console for level completion guidance.