"Fine weather today"

Europe is sweating.

That was the headline on the Norwegian news media TV2 this week. In Paris, temperatures reached 41 degrees Celsius, and crowds flocked to fountains throughout the city to cool down. The heat wave hitting Europe is yet another reminder that the climate is changing.

In this new reality, weather forecasters are tipped to play an important role.

Norwegians love talking about the weather, and we've got good reasons for it: Norway has long been a superpower in weather forecasting and meteorology. The Norwegian Vilhelm Bjerknes is regarded as the founder of modern weather forecasting, and it was he that, right after World War I, established the Bergen School (Bergensskolene), a movement within meteorology.

During World War II, Norwegian meteorologists played an important role for Britain, after the Germans restricted weather forecasting for civilians in Norway. Among the Norwegian meteorologists was Sverre Petterssen. It was his accurate weather forecasts that made D-Day take place on June 6th - a day the Germans thought the weather would be too bad for the British to invade.

For my part, I understood how strong Norway's influence on weather forecasting was back when I lived in London. After hearing that I was Norwegian, my rugby teammates mentioned the three things they knew about Norway: Ole Gunnar Solskjær, Henrik Ibsen, and the weather forecast app yr.no.

No more surprises

Weather forecasting has always been notoriously uncertain - that's just how it is. Until now. Breakthrough after breakthrough is happening that could ensure we're never again caught in the middle of downpours we never saw coming. If you don't have an umbrella going forward, it's your own fault.

The companies developing weather forecasting services are already tech-beheamoths: Google DeepMind, with GenCast, and Huawei with Pangu-Weather, have both created AI models that can predict the weather. But it's Microsoft that's currently winning the weather forecast race.

The Bill Gates company has launched a model called Aurora, and it can predict 10-day weather forecasts on a smaller scale than competitors. Simply put, this means Aurora can "see" the weather with a bigger magnifying glass than other AI models. This makes the forecasts more accurate for exactly the area you're trying to get weather data for.

More than just weather

But unlike the other models, Aurora isn't just a weather forecasting machine. It's built as a foundation model - not unlike ChatGPT. This means it can be adapted to solve various tasks that use weather data.

Among other things, Aurora can predict the amount of air pollution in an area, estimate waves, and warn about tropical storms. But what makes the model special is that it doesn't just understand the atmosphere; it understands the entire Earth system, and that opens up possibilities. According to the New York Times, a startup has already trained the model to predict how the renewable energy market will develop going forward.

Happening fast

The European Centre for Medium-Range Weather Forecasts (ECMWF) is the world leader in forecasting global weather conditions weeks ahead. In February, they implemented the world's first fully operational weather forecasting system powered by artificial intelligence.

The new AI forecasts are far simpler, faster, and cheaper to produce than traditional weather reports. In most cases, they're also more accurate. Where we previously would need large teams of experts and enormously expensive supercomputers, you can now create the same weather forecasts on a regular laptop.

This innovation opens doors for poorer countries that can't afford their own weather forecasting services or powerful computers to use such services. Now they can create their own weather forecasts much cheaper and easier. And this is a major breakthrough, given that weather forecasting is critically important for transportation, agriculture, and industry, among other things.

A month later, researchers from The University of Cambridge published an article about their model Aardvark Weather. They say their model will be able to drastically reduce costs and make the development of customized weather models accessible.

The weather forecasting systems we all rely on have been developed over decades, but in just 18 months, we’ve been able to build something that’s competitive with the best of these systems, using just a tenth of the data on a desktop computer. - Richard Turner, professor at Cambridge Department of Engineering, and weather forecasting researcher at the Alan Turing Institute.

A turning point

But no AI model works without datasets. Dr. Perdikaris was involved in the Aurora project. To the New York Times, he expresses concern about the Trump administration's cuts to agencies that handle weather data.

...the Trump administration’s cuts to agencies including the National Oceanic and Atmospheric Administration, the National Science Foundation and the National Weather Service could stymie further improvements in A.I. forecasting tools, because federal data sets and models are critical to developing and improving A.I. models. - Dr. Perdikaris.

Artificial intelligence is both frightening and extremely exciting. Sverre Petterssen made his weather calculations in what became a turning point in World War II. Now the meteorology field has taken new steps. Through Aurora, Microsoft has opened up the field and laid the groundwork for highly customised weather tools.

This could, unfortunately, prove to be invaluable in a world where temperatures are rising, heat records are being broken, and the climate crisis is getting less attention.

More Amazon robots

Amazon has spent years automating work tasks at their warehouses. Now Bezos has deployed more than one million robots at these workplaces, according to the Wall Street Journal. This is the highest number they've ever had and close to the number of human workers at the facilities.

The company now claims that 75 percent of all deliveries worldwide have robots involved.

Another batch of AI buddies enter the market

Research shows that many use chatbots for emotional needs, and that the interactions can sometimes prove problematic for people's mental health. You can read more about this in this Kludder post.

Last week, Anthropic, the company behind the chatbot Claude, could report that almost 3% of interactions are users trying to fulfill a psychological need such as seeking advice, having company, or engaging in romantic role-play.

But now a new AI chatbot has entered the scene. With its 100,000 monthly users, startup Portola has found a market: Young women. Through an app, you can build your own Tolan, a little alien blob that becomes your friend. But unlike other chatbots, Tolan is programmed to avoid sexual and romantic behavior. It also reminds you to do things you like, such as reading a book or going for a walk, and simply putting your phone away sometimes. If you want to read more about Portola, Tolan, and what they do differently, this article from Wired is worth reading.

Silly season has come to the AI industry

There's drama between OpenAI and Meta these days. I was a bit skeptical about including this story, as it falls more under gossip than an actual AI news story. But hey, we all need some drama in our lives!

Mark Zuckerberg and Meta are building a superintelligence lab, with the goal of achieving artificial general intelligence. To achieve this, they've been targeting OpenAI's most renowned AI researchers.

And it's clear that it pays off to be an AI expert: Wired, and several media outlets, report that the Facebook owner is offering total packages of up to 300 million dollars over four years to switch jobs. Not bad.

These statements are untrue - the size and structure of these compensation packages have been misrepresented all over the place - Meta spokesperson Andy Stone to Wired.