How AI Is Learning to Predict Our Extreme Weather

[00:00:00] 3, 2, 1. What will the world look like 10 or 20 years from now? The Wall Street Journal's Future of Everything podcast is here to give you a peek. And we can't wait to show you what's coming. Subscribe now. Welcome to Tech News Briefing. It's Tuesday, August 6th.

[00:00:22] I'm Zoe Thomas for The Wall Street Journal. Meal planning, appointment scheduling, and task delegating are all part of what's known as mental load. There are apps out there that can help, provided that managing the tech doesn't become another chore. Plus, artificial intelligence is learning to predict the weather.

[00:00:44] Scientists say AI-based programs could one day calculate forecasts faster and at a lower cost than existing methods. We'll tell you how they work and where things stand with their development. But first, there's the work you get paid for and the work that seems to occupy every

[00:01:05] other moment of your life. Shifts of cooking, cleaning, child care, and the endless to-do lists in our heads. They're called the mental load. A recent study of more than 2,000 parents commissioned by Skylight, the maker of a touchscreen

[00:01:20] calendar, found that parents devoted an average of 30 hours a week to managing family schedules and household tasks. If you're feeling overwhelmed by this, then our family and tech columnist Julie Jargent may be able to help. She's been looking into this cognitive burden and some possible tech solutions.

[00:01:39] Julie, you tested out three tools and suggest that they could help address some of the biggest pain points in mental planning. Let's take them one at a time. What about making all of those lists? There are multiple digital lists. There's one called Todoist, which is really popular.

[00:01:54] It's a task management app that started out as a workplace tool to have teams coordinate on things that they have to do to finish projects. There's a free version of the app. You just type in whatever to-do you have, and it gets added to a list.

[00:02:09] You can choose whether you want to complete that list today or at some future date. Then it'll send you reminders to do those. Then you can also delegate tasks to other people. You can break tasks into smaller steps, and you can set reminders for other people to

[00:02:25] complete their to-dos. What about meal planning, which I have to admit is something I always leave up to my stomach at the last minute? Is there an app to maybe help me stop doing that? There are a lot of apps.

[00:02:37] One that I looked into and that I like the way it was organized is called Plan to Eat. What's great about this one is you can import recipes from anywhere. You can just go to a website where you find a good recipe online, and you can import it

[00:02:51] to the app automatically. It'll just sort the ingredients out separately into a shopping list so you know exactly what to buy, and it's organized by the place in the grocery store where you would find those items. You can also import recipes from a cookbook.

[00:03:07] You can just take a picture and send that in or recipe cards if people still have those. And then you can also add those recipes to a calendar so that you can know what's for dinner each night.

[00:03:18] And you can just use the calendar in the app, or you can sync it with your own Google or Apple calendars. And what about dreaded cleaning chores? So I experimented with an app that I'd heard about from other people called Sweepy.

[00:03:33] And it's an app that is designed for younger kids. This is good for them because you can assign them chores on a room-by-room basis, and then you can specify exactly what you want them to do, whether you want them to make their

[00:03:45] bed or clean off their desk or what have you. And then they earn coins within the app that allow them to buy virtual decorations to decorate a virtual room. And kids like that kind of thing. The catch is you have to remind them to check the app.

[00:04:04] What are some of the all-in apps to manage things like family schedules and chores and shopping lists? Yeah, so there are some kind of one-stop shops that will do everything. There's Maple Cozy, and then there's the Skylight, which contains a calendar and you can manage chores on that.

[00:04:21] A lot of couples where stress is involved with a mental load is the sharing of these duties. And so if you're going to use an app, then you really need that commitment from your

[00:04:30] partner to also use that app and make sure that you don't have to remind them to do it. But if it becomes a chore in and of itself to manage the app and to have to remind your partner to participate, then it's just another chore.

[00:04:43] All right, that was our family and tech columnist Julie Jargon. Coming up, should you bring an umbrella? Companies are working on AI systems that can better predict the weather. We'll tell you about them after the break. What then will the future reveal?

[00:05:06] There's one thing we know about the future. It's being built now. We all have a stake in the future. The future. The future. The future. And the Wall Street Journal's Future of Everything podcast is here to give you a glimpse of what's on the way. I'm Danny Lewis.

[00:05:21] Join us as we dig into how science and technology are shaping the future. Oh, that is where you and I are going to spend the rest of our lives. Subscribe wherever you get your podcasts. Faster and more accurate weather predictions could help businesses, governments, and individuals

[00:05:44] plan for the future, especially as extreme weather has become more deadly and storms are more costly. Artificial intelligence is being trained to calculate those forecasts, and it could change the way we prepare for storms. Here now with more on this is our reporter Eric Kneeler.

[00:06:02] So Eric, how do meteorologists come up with weather predictions now? So right now if you can imagine there are these large scale computer programs that give forecasts for large sections of the globe, say North America, for example, or Europe or Asia.

[00:06:20] And these big computer programs take in a lot of data right now from weather balloons, from ships at sea, from satellites and so forth. And they make their predictions based on what are already well known, well understood equations.

[00:06:37] So running these programs takes a lot of data and a lot of computing power. And it's also very expensive to run these things. So what do AI models do differently? The weather AI models are looking for patterns in the atmosphere from previous data.

[00:06:56] So it takes a lot of time and effort to train these models, training on years of information without getting too technical. New AI weather models are looking for patterns from existing weather conditions that repeat things that happened in the recent past.

[00:07:14] And what are the benefits to AI weather predictions over the current methods? So if you just look at computing time, for example, that's one of the big expenses. Because the AI algorithms are not going back and solving equations like the supercomputers

[00:07:30] and the weather models are doing, the computer weather models, they're just hunting for patterns. It's much faster, up to thousands of times faster. And some of the big tech companies as well as even some smaller ones as well are starting

[00:07:44] to put together weather models that can do some amazing things. For example, Microsoft released a forecast tool called Aurora back in May that can produce a five-day global weather prediction for air pollution and a 10-day weather forecast that

[00:08:00] are 5,000 times faster than existing models run by NOAA and the European centers. What about accuracy? I mean, everybody wants to know, you know, is it actually going to rain tomorrow? Exactly. So one of the things they're working on is accuracy.

[00:08:15] Some of the bigger models, the ones that are coming out from Google DeepMind, Microsoft Aurora, NVIDIA is working with the weather company, for example. These models so far have done pretty well. They don't have the kind of refinement that we're looking for.

[00:08:31] They may show, for example, a big front moving across a large section of North America or big storms across England, but they're not going to give you the granular detail that's going to make a big difference for someone having to make business decisions or even

[00:08:47] whether or not you carry your umbrella that day. An AI forecasting model from Windborne Systems is being tested by the National Oceanic and Atmospheric Administration, or NOAA. How does it work and how did it affect the accuracy of NOAA's predictions?

[00:09:03] What Windborne is doing is collecting data with these weather balloons across large sections of the Pacific and Atlantic oceans, for example, and getting that information to NOAA. In a recent study, for example, they found that they could improve the accuracy of NOAA's

[00:09:18] own storm ground track forecast by up to 18%. The firm was also recently awarded contracts with the Navy and the Air Force to further develop its AI forecasting model to help pilots and ship captains understand better and get more warning for weather that the military needs to know about.

[00:09:39] I mean, early warning systems are really important. How are researchers developing AI programs to help with that? There's some interesting projects going on right now, one at the University of Oklahoma, for example. So, they're trying to look at patterns again, putting things together, and they're hoping

[00:09:57] to extend the warning time for hail and tornadoes from the current time of 15 minutes out to maybe even an hour. They're already doing this with hail, they hope to do the same thing with tornadoes. That would be a huge improvement, as you can imagine.

[00:10:13] When are these systems expected to become the standard for weather forecasting? Big weather agencies like NOAA, National Weather Service, the Europeans are already starting to evaluate them, how well they're working. The folks that I've spoken to at these agencies say it's probably going to be anywhere from

[00:10:34] two to five years before we see this as an actual part of your regular forecast. They need not just more accuracy, but more granularity. They need more detail in what's happening in the weather. So as the AI weather models and AI weather companies start getting more compute power,

[00:10:55] for example, faster processing speeds, more ability to run these AI models, that kind of accuracy is going to improve. All right, that was our reporter Eric Kneeler. And that's it for Tech News Briefing. Today's show was produced by Julie Chang with supervising producer Catherine Millsop.

[00:11:14] I'm Zoe Thomas for The Wall Street Journal. We'll be back this afternoon with TNB Tech Minute. Thanks for listening.