Huawei's AI weather model among top performers in China tests
Meteorological agencies worldwide are racing to develop and roll out new AI tools: the US National Hurricane Center teamed up with Google this month to improve tropical cyclone forecasts, while the premier European forecasting body became the first major prediction center to release a model this year. — Pixabay
China's weather agency is testing more than a dozen artificial intelligence models in an effort to enhance its forecasting, with a system from Huawei Technologies Co showing accelerated improvement.
The best models from the trial will be prioritized for deployment by provincial bureaus, and granted priority access to official weather data, according to the China Meteorological Administration, which is running the program. The CMA has said it wants to ensure "orderly and standardized development' as the technology rapidly develops at home and abroad.
In recent weeks, Huawei's Pangu-Weather has improved relative to others in the program. It was the second-best last month and over the first half of June, according to CMA data, compared with its seventh-place ranking in January. In particular, the model was better able to predict the intensity and geographic scope of several heavy rainfall events across southern China in April.
Huawei didn't respond to a request for comment.
Meteorological agencies worldwide are racing to develop and roll out new AI tools: the US National Hurricane Center teamed up with Google this month to improve tropical cyclone forecasts, while the premier European forecasting body became the first major prediction center to release a model this year.
While AI models have shown promise, outperforming traditional physics-based methods on a number of measures, they do have some limitations. Many are trained on historical data, which could make them prone to underestimating extremes as the Earth's warming climate enters uncharted territory.
China stands out for the number of models it's publicly testing. The trial scores them based on how well they predict key metrics such as temperature, wind speed, and humidity. Performance scores are updated daily, and detailed monthly evaluation reports are posted online.
Of the 14 models being evaluated, other top performers include Fuxi from Fudan University, and Puyun from Hangzhou-based startup Meta Carbon Co. Another is Sphere Fusion Forecast, jointly developed by China's state weather agency, Tsinghua University, and US chipmaker Nvidia Corp.
China's weather industry is closely regulated and regarded as a sector with important national security considerations, and only government meteorological agencies are allowed to issue public forecasts. New measures took effect this month regulating the use of AI, with the state weather bureau and the powerful Cyberspace Administration of China spearheading the initiative. – Bloomberg
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
The Star
10 hours ago
- The Star
Adobe brings AI-image generation app to phones, adds partners
FILE PHOTO: A logo of Adobe Inc. is pictured at the company's office in Citywest Business Campus, Saggart, Ireland October 19, 2021. Picture taken October 19, 2021. REUTERS/ Tom Bergin/File Photo SAN FRANCISCO (Reuters) -Adobe Inc. released its first dedicated artificial intelligence smartphone app on Tuesday that includes AI models from the company and partner firms, in a bid to tap into a growing trend of sharing AI images and videos over social media. The new app, called Firefly, packages Adobe's own AI model together with models from new partner firms Ideogram, Luma AI, Pika and Runway, and is available on iOS and Android phones. Previously, the service was available only as a web version, that used models from Alphabet's GOOGL.O> Google and OpenAI in addition to Adobe's model. Those models will also be available in the app, Adobe said. Generating images that can easily be shared on social media has become a key driver of AI interest, with OpenAI's Ghibli-style AI images driving record traffic to the ChatGPT creator. Adobe's mobile service will offer subscribers unlimited basic image generation from Adobe models, while it will charge extra for access to the company's premium models and those from its partners. The subscription cost will be the same as for the web versions of Firefly, which start at $10 per month. The San Jose, California-based company had earlier released AI tools along with the mobile app version of its popular image-editing program Photoshop. Adobe has not disclosed how much it pays the partner models on the Firefly app. The company had promised users that its AI model is trained only on material that it has a legal right to use, with Adobe offering protection against copyright claims. Ely Greenfield, Adobe's chief technology officer for digital media, said Adobe's approach has also gained some resonance among consumers. "Even for many of our individual customers, that promise of the commercial safety and the story about how Firefly is trained continues to be a really important differentiator," Greenfield said. (Reporting by Stephen Nellis in San Francisco; Editing by Harikrishnan Nair)
The Star
14 hours ago
- The Star
Volvo Group, Daimler look to technology joint venture in bid to save costs
FILE PHOTO: President and CEO of AB Volvo Martin Lundstedt stands in front of a Volvo truck, on the occasion of the publication of an interim report, in Gothenburg, Sweden, April 20, 2023. TT News Agency/Adam Ihse/File Photo GOTHENBURG (Reuters) -European truckmaking rivals AB Volvo and Daimler Truck hope to cut costs and reduce reliance on suppliers by jointly developing a software-defined vehicle program, they said on Tuesday. Fleets and truck manufacturers, like automakers, have been racing to produce technology-packed vehicles while also grappling with the need to reduce costs. Truckmakers are currently heavily reliant on suppliers because their software is closely tied to hardware, but Daimler and Volvo's new business - called Coretura - aims to develop a software-defined vehicle platform and reduce that dependence. The companies are looking to create an "industry standard", Daimler Trucks CEO Karin Radstrom told reporters. "We're looking at how we can move from our current reality where we are very much dependent on our suppliers - which drives both cost and sometimes timelines - and instead looking at what's the next generation of software that we need to bring to the vehicle", Radstrom said. The Gothenburg-based venture will employ 50 employees to start, with the hope of first deliveries of its connectivity platform in 2027 and further deliveries towards the end of the decade. "Everything in the vehicle industry is very much oriented around software and controlled by software", said Johan Lunden, a Volvo veteran and the newly appointed CEO of Coretura. Software will play an increasingly vital role in achieving sustainability, productivity, and safety targets in the future, he added. While rivals, Volvo Group and Daimler have collaborated on various businesses in recent years, such as within charging and hydrogen fuel cell development. (Reporting by Marie Mannes in Gothenburg and Jesus Calero in Gdansk; Editing by Matt Scuffham)
The Star
14 hours ago
- The Star
Huawei's AI weather model among top performers in China tests
Meteorological agencies worldwide are racing to develop and roll out new AI tools: the US National Hurricane Center teamed up with Google this month to improve tropical cyclone forecasts, while the premier European forecasting body became the first major prediction center to release a model this year. — Pixabay China's weather agency is testing more than a dozen artificial intelligence models in an effort to enhance its forecasting, with a system from Huawei Technologies Co showing accelerated improvement. The best models from the trial will be prioritized for deployment by provincial bureaus, and granted priority access to official weather data, according to the China Meteorological Administration, which is running the program. The CMA has said it wants to ensure "orderly and standardized development' as the technology rapidly develops at home and abroad. In recent weeks, Huawei's Pangu-Weather has improved relative to others in the program. It was the second-best last month and over the first half of June, according to CMA data, compared with its seventh-place ranking in January. In particular, the model was better able to predict the intensity and geographic scope of several heavy rainfall events across southern China in April. Huawei didn't respond to a request for comment. Meteorological agencies worldwide are racing to develop and roll out new AI tools: the US National Hurricane Center teamed up with Google this month to improve tropical cyclone forecasts, while the premier European forecasting body became the first major prediction center to release a model this year. While AI models have shown promise, outperforming traditional physics-based methods on a number of measures, they do have some limitations. Many are trained on historical data, which could make them prone to underestimating extremes as the Earth's warming climate enters uncharted territory. China stands out for the number of models it's publicly testing. The trial scores them based on how well they predict key metrics such as temperature, wind speed, and humidity. Performance scores are updated daily, and detailed monthly evaluation reports are posted online. Of the 14 models being evaluated, other top performers include Fuxi from Fudan University, and Puyun from Hangzhou-based startup Meta Carbon Co. Another is Sphere Fusion Forecast, jointly developed by China's state weather agency, Tsinghua University, and US chipmaker Nvidia Corp. China's weather industry is closely regulated and regarded as a sector with important national security considerations, and only government meteorological agencies are allowed to issue public forecasts. New measures took effect this month regulating the use of AI, with the state weather bureau and the powerful Cyberspace Administration of China spearheading the initiative. – Bloomberg
