"Google says it spent five months fine-tuning its latest agent to make it capable of seamlessly navigating and completing multi-step tasks across some of the most popular phone apps in use today. As you might imagine, Google claims the system is capable of some significant feats of automation. For instance, the company says the agent can read a class syllabus in Gmail and then put all the books you might need for that course in a shopping cart."
"Google adds the system is even more powerful when it can pull context from your phone's screen or an image. The company paints a scenario where users might see a travel brochure in a hotel and ask Gemini to find a similar tour on Expedia. Understandably, some people might be reluctant to give an AI agent control over their phone."
"Google says Gemini Intelligence won't begin working on a task until instructed to do so by a user. Moreover, any task that involves Gemini buying something on your behalf will require you to confirm the purchase. Users can also decide when Gemini can access their data using Google's familiar permissions menu, and a progress bar allows users to stop Gemini at any time."
"Google plans to first bring Gemini Intelligence to recently-released Pixel and Samsung Galaxy phones. It will be interesting to see how much usage people get out of the system. It's not like most apps are hard to use; if anything, over two decades of mobile design, the majority have been streamlined to the point most of us can complete tasks like calling an Uber without much thought."
Gemini Intelligence is an agent system for Android-based phones that automates tedious, multi-step tasks across popular apps. Google says it fine-tuned the agent over five months to navigate and complete tasks seamlessly. The system can interpret information from Gmail, such as reading a class syllabus and adding required books to a shopping cart. It can also use context from what appears on the phone screen or from an image. For example, a user can view a travel brochure and ask for a similar tour on Expedia. The agent only starts when a user instructs it, requires purchase confirmation, and can be stopped at any time through a progress bar and permissions controls.
Read at Engadget
Unable to calculate read time
Collection
[
|
...
]