Top 5 This Week

Related Posts

DeepMind’s latest trick? Predicting the weather

After mastering Go and StarCraft, DeepMind is taking its AI into another challenging arena: predicting the weather. The Alphabet-owned company has been quietly working with the Met Office over the past few years, and today, they report the fruits of their collaboration in the journal Nature. In short, DeepMind has devised a new machine learning model that can predict whether it’s going to rain within the next couple of hours. 

This type of weather forecasting is known as precipitation nowcasting: predicting rainfall on a very short timescale, up to two hours before a downpour. Today's weather forecasts are pretty nifty at predicting rain further ahead in the future, from six hours to about a couple weeks ahead. But any sooner than that is where blind spots appear, and that’s where machine learning can bridge a much-needed gap.

A deluge of rain isn’t just an annoyance for someone who just got their hair done. Being able to foretell heavy rain in advance is crucial for everyday but important situations such as road safety, air traffic control and early-warning systems for flooding. The unfolding climate crisis also means that extreme weather events, such as heavy rainstorms or flooding, will only become more frequent. The ability to predict rainfall better and faster is pretty important for making quick decisions in these situations: halting a train or evacuating a building, for example.

The Met Office relies on radar imagery to predict when the heavens will open up. Radar works by sending a beam into the atmosphere, and then timing how long it takes to reflect, which tells you how much moisture there is in the atmosphere. The more moisture there, the more rain there will be. The data are then sent to the Met Office HQ, where they are processed to get a picture of the precipitation over the UK. DeepMind’s model was trained on radar imagery from the UK between the years 2016 and 2018, to then be able to reliably predict what will happen in an hour or two into the future. 

The team of researchers also ran their model past over fifty meteorologists working at the Met Office. In the blind study, the meteorologists were asked to compare DeepMind’s approach, dubbed DGMR, with another nowcasting method known as PySTEPS, as well as a different deep learning method. DeepMind’s model ranked first over the others in terms of accuracy and usefulness in almost 90 per cent of cases.

Other scientists are not as bowled over by the results. “I do not see any revolution in forecasting here,” says Peter Clark, a meteorologist at the University of Reading. And relying on this metric to prove the usefulness of their model puzzles him. “I remain surprised that they did not choose to use a more appropriate objective score,” he says. “Little is said in detail about how the assessments were carried out, or even what was actually assessed.” DeepMind refrains from putting a concrete number on how much more accurate its model is compared to the other existing models. “The approach that we wanted to take is this more measured approach, rather than reporting a [single] number,” says Shakir Mohamed, a senior scientist at DeepMind and author on the paper. 

Most PopularSecurityHackers Found a Way to Open Any of 3 Million Hotel Keycard Locks in Seconds

Andy Greenberg

Backchannel8 Google Employees Invented Modern AI. Here’s the Inside Story

Steven Levy

ScienceThe Keys to a Long Life Are Sleep and a Better Diet—and Money

Matt Reynolds

GearThe Omega x Swatch Snoopy MoonSwatch Has Landed

Jeremy White

Rather than completely shaking up rain forecasting as we know it, DeepMind’s research might simply offer a different approach. “It performs similarly with other current cutting-edge type models,” says Rob Thompson, a meteorologist at the University of Reading. But their model isn’t miles ahead; “it’s possibly slightly better,” he says. David Schultz, a meteorologist at the University of Manchester, agrees. “The proposed scheme seems to beat those.  Game changer? I don’t know. Big improvement over some existing methods in their paper? Yes.”

There aren’t any immediate plans to put the model into use, says Mohamed, but the team hopes to eventually use the model to inform future weather reports for real. But AI will not one day replace weather forecasters, the researchers say. “You are going to need experts and humans in the loop in order for them to make sure what they’re understanding in terms of the predictions are sensible, and then how to communicate that to the public,” says Suman Ravuri, a research scientist at DeepMind and also an author on the paper. “It’s hard to see an automated system doing all of that with the requisite nuance and skill.”


More great stories from WIRED💼 Sign-up to WIRED’s business briefing: Get Work SmarterThe race to stop fish becoming the next factory farming nightmareWhat to do if your Facebook account is hackedMicrosoft is heading for a new antitrust showdownHow Out Run changed video games foreverThe draconian rise of internet shutdownsA radical plan to treat Covid’s mental health falloutThe 100 hottest startups in Europe in 2021🔊 Subscribe to the WIRED Podcast. New episodes every Friday

This article was originally published by WIRED UK

Popular Articles