At I/O 2024 in May, Google showed two examples of agent experiences accessible through Gemini. Google could be ready to share details with Project Jarvis this December about an agent that runs on Chrome and is powered by Gemini 2.0.
“I suppose [agents] As an intelligent system that exhibits reasoning, planning, and memory. We can think multiple steps ahead and work across software and systems to get things done on your behalf and most importantly, under your supervision. ”
—Sundar Pichai talks about AI agents
According to The Information, Google is “developing artificial intelligence that can take over your web browser to complete tasks such as collecting research results, purchasing products, and booking flight tickets.”
Project Jarvis, named after JARVIS from Iron Man, runs on Google Chrome and is intended for consumers (rather than businesses) to “automate everyday web-based tasks.” The article doesn’t specify whether this is for mobile or desktop.
At I/O, Pichai showed how “Gemini and Chrome work together to help you organize, reason, compose, and do a variety of other preparation tasks on your behalf.” This staged scenario typically occurred via gemini.google.com with no other UI visible compared to the previous example that occurred via Gemini for Android.
When given a command/action, Jarvis says, “frequently take a screenshot of what’s on your computer screen and then take a screenshot of what’s on your computer screen before performing an action, such as clicking a button or typing in a text field.” Work by “interpreting the shot.” In today’s report, Jarvis says that “the model is relatively slow because it has to think for several seconds before performing each action.” So this likely won’t work on your device yet and still requires the cloud.
Jarvis is said to be powered by Gemini 2.0 and could be previewed “as early as December,” another rumor confirmed yesterday. Jarvis may then be available to early testers, so a launch doesn’t seem imminent. It makes sense for Google to have examples of flagship products leveraging Gemini 2.0. Past model releases have done something similar, but Jarvis seems to have gotten more specific.
FTC: We use automated affiliate links that generate income. more.
