The first satellite with AI on board is now in orbit,...

• This week the European Space Agency (ESA) and Intel announced that they had successfully launched PhiSat-1, the first satellite with integrated AI processing.
• The PhiSat-1 uses the Movidius Myriad 2 chip from Intel, which was not originally developed for space travel.
• PhiSat-1’s integrated AI can select photos of the earth and automatically delete them if they are covered by clouds.
• AI has become mainstream on Earth, but getting it on satellites was a huge challenge and PhiSat-1 could pave the way for innovation in how satellites detect natural disasters or even communicate with Mars rovers.
• You can find more stories on the Business Insider homepage.

On September 2, a satellite the size of a cereal box was launched into space.

Called the PhiSat-1, it was intended to monitor polar ice and soil moisture, which – at least on the surface – makes it a pretty nondescript part of the kit.

For the developers of the satellite – the European Space Agency (ESA), chip giant Intel and Irish robotics company Ubotica – this launch meant months of work and was postponed by a failed rocket launch, two natural disasters and a global pandemic.

It was also a major technological leap forward.

PhiSat-1 is the first satellite to go into orbit with AI on board – and the technology could transform the way we respond to disasters like oil spills and forest fires.

Why did it take so long to get AI into space?

ESA’s Gianluca Furano told Business Insider that while AI is popular for building products on earth, getting the technology onto satellites has been a major challenge.

“Using AI in a data-critical application is not easy, even on Earth,” he said. Flying an AI chip like the one aboard the PhiSat-1, 329 miles above the planet, means repairs, software patches, or upgrades are all the more difficult.

“The complexity just explodes,” said Furano.

In addition, radiation in space makes it difficult to build an AI chip that can withstand orbit.

“Whatever silicon contains is disrupted by ionizing radiation,” said Furano. This means that computer chips that would function perfectly on Earth have unusually high error rates and could even catch fire if sent into space without shielding or modification.

While many satellites carry custom-made space-grade chips, PhiSat-1 uses Intel’s Myriad 2 chip, a commercially available chip found in a handful of commercial drones – and even the headset for Magic Leap’s AR glasses.

Ubotica took the Myriad 2, adjusted its software, and built electronics that can shut it down if it’s likely to overheat. The researchers tested it by taking it to the largest particle accelerator on earth at CERN in Switzerland.

The satellite was ready by May 2019, but it took longer than planned to get into space. The original launch was scheduled for September 2019, but the rocket that was supposed to pick up PhiSat-1 failed to launch, leading to an investigation.

Meanwhile, the coronavirus pandemic set in – and to top it off, two hurricanes hit the launch complex in French Guiana and a ground station in South Korea.

On September 2, PhiSat-1 finally made it into space and has since been successfully in orbit to put its AI into practice.

With PhiSat-1’s AI, clouds can be ignored – which is far more impressive than it sounds

Like the satellite itself, the PhiSat-1’s built-in AI function doesn’t seem glamorous at first, as photos of the earth it thinks are too cloudy are automatically deleted.

Before PhiSat-1 sends back images, its AI decides whether they are clear of cloud cover. If the AI ​​decides that more than 70% of a photo is obscured by a cloud, that image will automatically be scrapped.

This is useful because Phisat-1 saves a lot of energy. Aubrey Dunne, Ubotica’s chief technology officer, told Business Insider that 67% of the earth’s surface is normally covered with clouds. This means that the AI ​​will crop many images instantly instead of using processing power and sending useless images back to scientists.

“Cloud discovery is about data reduction,” he said.

This could pave the way for fire fighting and Mars rover driving

Furano said that while Phisat-1’s successful launch with its cloud scrapping AI is exciting, current uses of the technology are “low hanging fruits.”

“The really exciting part comes when you think of doing onboard processing with machine learning [the type of AI PhiSat-1 uses] on applications where you need a very quick response, “he said.

As an example, Furano said that AI could massively accelerate satellite alerts of disasters like oil spills.

Ubotica’s Aubrey Dunne said the company was already working on adapting the technology to build satellites that could automatically detect forest fires and oil refineries’ flares. “We’re trying to reduce latency, so we’re trying to use AI to run applications where time is really important to respond and fire detection is a good example,” he said.

“You want to try to inform authorities and relevant people on the ground in the relevant places about the location and extent of the fire and how it is changing, moving and moving without having to wait a day for the data to be downloaded and one day having to wait another day when it can be processed on site, “he said.

Dunne believes AI could also have uses outside of the world, such as speeding up communication with rover robots on Mars.

Ubotica aims to launch the next Myriad AI chip into space in early 2022. “There are some satellites that are about twice the size of the PhiSat satellite, and they are partly funded by ESA, but also partly financed commercially, including the commercial side,” said Dunne, although he did not explain where the commercial came from Funding would come.

Jonathan Byrne, head of Intel Movidius technology office, told Business Insider that Intel sees AI satellite chips as an emerging market but one that has the potential to grow.

“The technology is 20 years old and is on satellites. 20 years ago GPUs grew massively – and you see the same thing, ”said Byrne.

Furano said the technology that is currently in orbit is 15 to 20 years behind the computing power found in smartphones today, meaning new, improved chips are high on its list. “We urgently need AI in space,” he said.