pocket-lint.com

Google launches Gemini 2.0, ushering in the era of AI agents

Summary

Google introduces Gemini 2.0, its new advanced AI model, promising advancements in multimodality and agent capabilities.

Project Astra is an experimental universal AI assistant powered by Gemini 2.0, promising better dialogue and understanding of your surroundings.

Project Mariner is an experimental Chrome extension that uses AI to complete web tasks.

Google has pulled the curtain back on Gemini 2.0, its most powerful AI model yet. It promises advancements in image generation, live conversations, and AI agents. Gemini 2.0's Flash model is available to Gemini and Gemini Advanced users on the web app, and is expected to land on mobile devices early next year.

"2.0 Flash also comes with new capabilities. In addition to supporting multimodal inputs like images, video and audio, 2.0 Flash now supports multimodal output like natively generated images mixed with text and steerable text-to-speech (TTS) multilingual audio." Google said in a press release.

Project Astra and Project Mariner are two new experimental projects powered by Gemini 2.0 which are currently being tested by "developers and trusted testers" that could make their way to products in the near future.

google-search Related

Google's Year in Search 2024 is here, and the results might not be what you expect

Google has released its Year in Search 2024, giving us a glimpse at the search terms that trended the most this year.

The agentic era is here

Project Astra is a new experimental AI assistant powered by Gemini 2.0

In addition to Gemini 2.0's Flash model, Google is also developing two experimental Gemini projects, Astra and Mariner.

Project Astra, Google's latest universal AI assistant, uses Gemini 2.0 to enable its agentic capabilities. It isn't available to the public yet, but in a video demo, Google demonstrates how it can be used as an everyday personal assistant. The demo shows Astra being used for tasks such as scanning the tag on your clothes to figure out how to wash them properly, scanning landmarks to find out more information on it, and asking it for recommendations based on data you provide via voice, text, screen share or a photo.

Astra promises better dialogue and conversation with the user by supporting multiple languages and better understanding of accents and uncommon words. It takes full advantage of Gemini 2.0 by leveraging Google Search, Lens, and Maps to increase its functionality as a personal AI assistant. Astra seems extremely promising, and I can quickly see how I could use it in the real world, especially while traveling to learn more about my destination's culture and language.

Google's Pixel 9 Pro XL camera Related

Google is bringing back this beloved Pixel Camera feature in a welcome update

Google giveth, and Google taketh away: Manual controls are making a comeback on your Pixel's viewfinder.

Project Mariner lets AI take action for you

It's an experimental Chrome extension that can complete tasks on the web for you

Next up is Project Mariner, which works in the browser as an experimental Chrome extension that uses AI to complete actions on the web for you. This is only available to trusted testers at the moment as it's in its early stages, but Google did release a demo showcasing how it works.

In the demo, the user prompts Mariner to research companies on a Google sheet for them. Mariner reads the spreadsheet, and begins researching the companies on the web to find their contact information and fills it in on the spreadsheet. You are in full control of the agent, so you can stop it any time. Mariner's user interface also explains what it's doing every step of the way. Google says it is working with its testing partners to make Mariner faster and smoother.

Project Mariner is a fascinating example of how AI on computers can start doing basic tasks for us, but whether we want it to is another question entirely. Based on Mariner's demo, you just sit there and watch it do all the work for you instead of it running in the background. To me, it still seems like it will be quicker just to Google and find out something yourself rather than watching AI do it for you.

Google is being cautious with the rollout of its new AI technologies, taking safety and security seriously and performing "extensive risk assessments" and testing for its Gemini AI initiatives.

ai-robot-1 Related

Google's new Genie 2 AI model is here, and it wants to suck the soul out of game design

Google's DeepMind team of scientists and engineers have just announced Genie 2, which can auto-generate entire 3D video game worlds on the fly.

Read full news in source page