TECHNOLOGY

Google's AI Upgrades: What's Next for Gemini Advanced?

USAWed Feb 19 2025
Google recently sent out a newsletter to its Gemini Advanced subscribers, hinting at exciting developments in the coming months. The subscription, priced at $19. 99 per month, is being promoted as a "pass to Google's next-gen AI. "This suggests that the updates might not be limited to just the Gemini app. The newsletter starts by looking back at recent experimental model launches. The first model mentioned is Gemini 2. 0 Pro Experimental. This model is designed for better factuality and exceptional performance in coding and math tasks. It aims to make complex tasks easier and more accurate. The second model is Gemini 2. 0 Flash Thinking. This model is built for speed and performance, and it shows its thought process in real time. This means users can see why it responded a certain way, its assumptions, and trace the reasoning. The newsletter also highlights last week’s launch of recall in the Gemini app and NotebookLM Premium. This is a big deal because it shows that Google is constantly improving its AI tools. The newsletter also teases some exciting updates coming in the future. One of the updates is about new ways to create with leading video, image, and audio generation tools. The Gemini app already provides access to Imagen 3, which was last updated in January. Veo 2 is behind a Google Labs waitlist. Audio generation is presumably referring to MusicLM and/or Lyria. It would be interesting if Gemini Advanced provided priority access to those standalone or Gemini-integrated tools. Another update is about being more productive with agentic tools that can take action on your behalf. This means you can spend your time how you want. In terms of agentic tools there’s Project Mariner, which CEO Sundar Pichai previously said would come to the Gemini app in some form this year. Other agentic features could be in Google Workspace, with I/O 2024 featuring a demo about how you could “ask Gemini to automatically organize email attachments in Drive, generate a sheet with the data and then analyze it with Data Q&A. ” The last point is more straightforward with Gemini 2. 0 Pro eventually hitting stable and becoming the default model for Advanced subscribers. It will be interesting to see how Gemini 2. 0 Flash Thinking gets presented to end users.

questions

    What are the key differences between the Experimental models and the stable models in terms of performance and usability?
    What measures will Google take to ensure the security and privacy of user data with the new features?
    Could the agentic tools be used to manipulate user behavior without their knowledge?

actions