Tech News

Google already allows you to detect forest fires in real time and predict how they will spread

(photo: Google AI)

At a press event focused on the artificial intelligence today in new york Google announced that it is taking its system forest fire detection powered by AI to USA Canada, Mexico and parts of Australia.

[article_mb_code]

It’s one of several “AI for good” efforts the company detailed this morning, which also included Google’s efforts to expand flood forecasting to more regions around the world.

The previously announced system uses models of machine learning trained on satellite data to track fires in real time and predict how they will develop. The feature is initially focused on helping first responders determine the best way to control the fire.

[article_mb_code]

Google says wildfire detection is also available in Brazil, Colombia, Sri Lanka, Burkina Faso, Cameroon, Chad, Democratic Republic of the Congo, Ivory Coast, Ghana, Guinea, Malawi, Nigeria, Sierra Leone, Angola, South Sudan. South, Namibia, Liberia and South Africa.

Google. (photo: Composition/Jose Arana/Google)
Google. (photo: Composition/Jose Arana/Google)

Google wants to help homes damaged by natural disasters

[article_mb_code]

Google also noted that it is using machine learning models to identify damaged buildings after natural disasters like hurricanes.

“By analyzing the same satellite imagery as their wildfire detection tools, the models provide crisis responders with information on the worst-affected areas in the first phase of the response effort,” he says. Google.

Google first applied the technology in partnership with the nonprofit organization GiveDirectly to identify and channel donations to low-income households affected by Hurricane Ian. Most recently, the company applied the models to support government relief efforts during recent flooding in Pakistan.

[article_mb_code]
Aftermath of Hurricane Ian in Florida. (photo: REUTERS/Marco Bello)
Aftermath of Hurricane Ian in Florida. (photo: REUTERS/Marco Bello)

Robots could generate their own code

New research demonstrated at the same Google event puts forward the notion of letting systems robotic write your own code. The concept is designed to save developers the hassle of having to go in and reprogram things as new information emerges.

The company notes that existing research and trained models can be effective in implementing the concept. All of that work can be critical in developing systems that can continue to generate their own code based on objects and scenarios found in the real world. The new work shown today is Code as Policies (CaP).

(photo: Google)
(photo: Google)

“With Chap, we propose to use language models to directly write robot code. Our experiments demonstrate that code output led to better generalization and task performance rather than direct robot task learning and natural language action output. CaP enables a single system to perform a variety of complex and varied robotic tasks without task-specific training.” Jackie Liang of Google.

The system, as described, also relies on third-party libraries and APIs to better generate the right code for a specific scenario, as well as support for languages ​​and emojis. The information accessible in these APIs is one of the current limitations.

Robots that generate their own code. (photo: Google)
Robots that generate their own code. (photo: Google)

The researchers note: “These limitations point to avenues for future work, including extending visual language models to describe low-level robot behaviors (e.g., trajectories) or combining CaP with exploration algorithms that can aggregate autonomously to the set of control primitives.

As part of today’s announcement, Google will release open source versions through your site github to build on the research that has been presented so far.

Back to top button