Important points
- A new report sheds light on Google’s next AI agent, codenamed “Project Jarvis.”
- Project Jarvis can automate web-based tasks such as booking flights and will reportedly be powered by Gemini 2.0.
- Google plans to make Project Jarvis official in December, but its rollout may be limited to a few people for testing purposes.
Google has big ambitions for AI, as evidenced by the numerous tweaks and upgrades it has made to its Gemini chatbot over the past few months. During the I/O developer conference in May, the company spoke briefly about “.A versatile AI agent useful in daily life” Google said some of this AI agent’s capabilities could be introduced to Gemini this year. A new report over the weekend revealed some new details about Google’s plans for this AI agent.
related
Google’s Project Astra brings Gemini AI to the real world
The idea for Google Glass is evolving with interruptions
According to an exclusive report by The Information (paywall), the project in development is likely going by the codename. project jarvis — Leverage your web browser to perform one of many tasks, such as booking a flight, researching information, or purchasing a product (via The Verge). Google plans to introduce Project Jarvis in December with a customized experience for Google Chrome, the report claims.
The timing couldn’t be better as it is powered by Gemini 2.0 and is scheduled to be released by December. Google wants to roll out the AI agent’s capabilities to a small number of users initially for testing, so it doesn’t expect Jarvis to have widespread access once it’s officially introduced. It’s also worth remembering, as The Information points out, that there is no set release schedule for December, and Google may choose not to show off Jarvis and its features before then. be.
So how does it work?
Based on the publication’s reporting, Project Jarvis is designed to improve “routine tasks” by continuously capturing and understanding screenshots of the screen before the user taps a text box or button to provide additional input. It is designed to automate web-based tasks. However, it is noted that the response is currently a bit slow because “the model has to think for several seconds before performing each action.” This suggests that Jarvis may not be ready for primetime appearances yet.
The newspaper said it spoke to three people with direct knowledge of the matter, but there are no images or videos showing how Jarvis works. But as our own Will Sattelberg pointed out in our coverage of I/O 2024, this AI agent looks like this:A version of the functionality promised by Humane and Rabbit on dedicated hardware.December is just around the corner and we hope to learn more about Project Jarvis and see what it can do in the coming weeks.
