Huawei builds AI to explore the universe
The company’s tech is powering an ambitious project to understand gravity and explore the birth of galaxies.
The barren landscape will be dotted with 130,000 antennas to read signals from tens of light years away. The Square Kilometre Array (SKA) Observatory will study gravity, and the birth of stars and galaxies. “Astronomy is about looking back in time by detecting signals that have been traveling through the universe,” said said Philip Diamond, Director General of SKA.
The project is supported by 13 countries across the globe and is working with Huawei to develop new advancements in artificial intelligence that will analyse the massive volumes of data detected in outer space. Diamond discussed this vision together with Huawei’s chief executive at the company’s annual summit in Shanghai last week.
Space exploration with AI
SKA will have two facilities in South Africa and Australia to gather and process huge amounts of data from the universe. Artificial intelligence will be “invaluable” to speed up the analysis of these oceans of data, Diamond said. “Our partners here at the Shanghai Astronomical Observatory have already been working with Huawei in this area, applying machine learning techniques to astronomy problems, such as pulsar searches and radio galaxy detection using simulated SKA images.”
The first phase of the project, which will begin construction in 2021, could produce five times the 2015 global internet traffic and be as powerful as two million computers. To handle these huge amounts of data, SKA will build alliances with regional data centres, Diamond added.
Diamond believes that the Observatory’s work will lead to new developments in AI, in areas like data visualisation and machine learning. “China is strongly involved in the project and has actually been a key partner since its inception,” he said.
Faster AI training tech
From astronomy to climate change and autonomous driving, every new area of human advancement will require insights from data with AI. Huawei wants to make this technology easier to use, more accessible and speedier.
One of the most intensive stages of developing AI is training the software to recognise things. For instance, training an algorithm to recognise a cat requires a metric ton of computing power. “The industry doesn't have nearly enough computing power to meet demand,” said Ken Hu, Huawei’s Deputy Chairman. The company is building stronger computing to increase the speed of complex AI model training and making it more affordable.
At Huawei Connect 2019 last week, Hu launched the world’s fastest AI training cluster. “Atlas 900 is a powerhouse of AI computing, and it will bring new possibilities to different fields of scientific research and business innovation – anything from astronomy to oil exploration. For models that used to take several months to train, now Atlas 900 can handle them in seconds,” Hu said.
Take astronomy where it usually takes scientists 169 days of full-time work to find a celestial body with specific features. Huawei’s AI model can identify a specific star in 10 seconds, he added. “It’s revolutionary and will free up scientists’ time for more important work.”
Huawei will invest $1.5 billion in its developer programme to grow AI talent. “We want to expand the programme to five million developers and better enable our partners around the world to develop the next generation of intelligent applications and solutions,” Hu announced.
It will work with governments to build innovation hubs and incubators, he added. “These hubs will bring together partners across the ecosystem, where we can carry out application pilots, cultivate talent, and develop standards as a team.” In China, it has set up hubs in Beijing, Shanghai and Shenzhen, where both local and overseas companies can get access to the company’s technology.
All of this - an open ecosystem, strong talent, and reliable technology - will be critical to see through a project that could fundamentally change our understanding of the world we live in.