AI, Automation, And The Great British Climate
You may have noticed that we've had some snow this month. And, as is traditional, the entire country collapsed into a state of paralysis.
It's quite fun for us to have a snow-day every now and again, but snow also brings lot of problems in its wake. The recent Beast From The East saw supermarkets emptied, ambulances grounded, water mains burst, and the economy sucker-punched.
But what if we could use modern technology to limit snow's impact on Britain? In our last post, we talked about the possibility of getting gritting robots out onto our roads. This time, we'll turn to our favourite national pastime: guessing the weather.
Why is Britain's weather so changeable?
The UK's climate is uniquely unpredictable. Our peculiar position (surrounded by sea but right next to a large continent) makes our weather perhaps the most diverse and changeable in the world.
For starters, the skies above Great Britain are the meeting point of five major air masses, which tussle for supremacy all day, every day. Above this tropospheric battle ground travels the Jet Stream, which can have a significant effect upon weather patterns. Then there are our seas, which use the Gulf Stream and other ocean currents to throw a hefty spanner into the climactic works. All in all, we are slap bang in the middle of a meteorological free-for-all, meaning that our weather is capable of dramatic change at phenomenal speed. We frequently have all four seasons, twice, in the space of an hour.
These (sometimes literally...) lightning changes in weather make it very hard to predict what the British climate is going to do from one minute to the next, let alone weeks in advance.
Nations with more predictable meteorological patterns can prepare for snow weeks or even months before the first flakes start falling. Britain is lucky if it gets a 24 hour warning (although that doesn't stop the tabloids from confidently predicting apocalyptic blizzards from August to April). Understandably, councils are reluctant to pour resources into snow-management measures when 6 times out of 10 the climate will decide that - despite all patterns pointing to snow - it's actually going to be warm and sunny.
Predicting the British weather is a problem (and national obsession) which has occupied the nation for millennia. Many of us quite like the uncertainty (sudden, random sunshine feels like a wonderful surprise treat, and mapping out the weather to the second would take a lot of the fun out of things). But in the case of things like The Beast From The East, a little advance warning would help us to get our act together.
How can Artificial Intelligence help?
Because weather systems are complex, they require a degree of mental flexibility to comprehend. So the most effective weather-mapping algorithms are equipped with 'deep learning' systems - programs designed along the same lines as human neural networks. These algorithms can pick up and process enormous amounts of data from weather stations and satellites, 'learning' all the while about the weather systems they're 'studying'. And, unlike human meteorologists, they can do this 24/7, without rest.
When you've got vast amounts of data being constantly and quickly processed by unrelenting, unforgetting algorithms, patterns quickly start to become obvious. Even a climate at fickle and changeable as Britain's could be rendered predictable by a sophisticated AI with deep-learning capabilities.
The problem at the moment is that (sci-fi though it sounds) nobody can say exactly HOW deep-learning systems reach the conclusions that they do. Like human brains, they develop their own methods of cognition which - accurate though they may be - make scientists a bit wary. If you can't show the route the machine used to reach its conclusions, then, scientifically speaking, those conclusions are inadmissible. After all, no human scientist would be allowed to present a conclusion without the accompanying hypothesis, research, and discussion.
Should weather AIs with deep-learning capabilities ever enter the fraught arena of British weather forecasting, however, it's likely that we'd eventually be able to predict snowfall with greater accuracy, and further in advance. This would enable us to make reasonable preparations - ultimately saving time and money when the snow eventually did hit.