AI Enhances Flood Warnings but Cannot Erase Risk of Disaster

A view shows a flooded schoolyard in Bamako, Mali, September 23, 2024. (Reuters)
A view shows a flooded schoolyard in Bamako, Mali, September 23, 2024. (Reuters)
TT
20

AI Enhances Flood Warnings but Cannot Erase Risk of Disaster

A view shows a flooded schoolyard in Bamako, Mali, September 23, 2024. (Reuters)
A view shows a flooded schoolyard in Bamako, Mali, September 23, 2024. (Reuters)

When floods ripped through parts of Europe in September, the scale of the destruction took people by surprise. The intense rains should not have, because those had been predicted by sophisticated forecasting systems enhanced with artificial intelligence.

But forewarned did not mean forearmed. Though the rains were accurately predicted, the effects in the deluged areas were not - a fact that highlights the difficulties of dealing with ever more common extreme weather.

AI has supercharged weather forecasting, using a range of statistical tools to analyze years of historical data and predict patterns, and at a lower cost than traditional numerical weather predictions.

AI technology can create more specific predictions ahead of events such as urban flooding or in complex terrain such as mountainous areas.

For example, Google-funded GraphCast, a machine learning–based method trained directly from reanalysis data, was found to outperform traditional models. Reanalysis data relies on past forecasts rerun with modern forecasting models to provide the most complete picture of past weather and climate.

But there are still gaps in knowledge, in how the information is used and in investment to strengthen data gathering models, experts say.

"In some cases and for some variables, AI models can beat physics-based models, but in other cases vice versa," said Andrew Charlton-Perez, professor of meteorology at the University of Reading in the UK.

One issue is that the effectiveness of an AI model is only as good as the information it is fed. If there is little input data, or extreme events happen more frequently at different times of the year or in different regions, weather disasters become more challenging to predict.

"A good use of the AI-based weather forecasts would be to complement and enhance our forecasting toolbox, perhaps by allowing us to produce larger ensembles of forecasts that enable accurate assessment and interpretation of the likelihood of extreme events," Charlton-Perez added.

COMMUNICATION IS KEY

Since January, the European Center for Medium-Range Weather Forecasts (ECMWF), an independent organization that provides predictions four times per day to European countries, has been using the Artificial Intelligence/Integrated Forecasting System (AIFS).

This data-driven forecasting model makes multiple predictions rapidly and delivers long-term forecasts of weather events like cyclones and heatwaves.

The ECMWF readings ahead of the September floods were accurate, experts say.

Thomas Wostal, press officer for meteorological observatory GeoSphere Austria, told Context/the Thomson Reuters Foundation that their numerical models - including the ECMWF's predictions - foresaw 300-400 millimeters (11.8-15.7 inches) of rain locally, which came to pass.

But even with accurate forecasts, scientists say communication is key, especially in an era when climate change means extreme weather is becoming more frequent.

"I think what happened with (the recent floods) ... is that it's so rare - a one in 150- to 200-year event - that even if the weather models capture it, there's a reasonable degree of uncertainty," said Shruti Nath, a postdoctoral research assistant in predicting weather and climate at Oxford University.

"You have to produce the warning in a way that is communicative, in the degree of severity it could possibly have on people, then people could see the cost of inaction versus the cost of action is actually much greater. So then they would actually put (in) more resources," she said.

EUROPE BEHIND THE CURVE?

Europe faces urgent climate risks that are outpacing policies and adaptation actions, a report from the European Environment Agency has warned.

Extreme heat, drought, wildfires and flooding will worsen in Europe even under optimistic global warming scenarios and affect living conditions throughout the continent, the EEA says.

After the floods, the European commissioner for crisis management, Janez Lenarčič, said the disaster was not an anomaly.

"These extreme weather events that used to be once in a lifetime are now an almost annual occurrence. The global reality of climate breakdown has moved into the everyday lives of Europeans," he said.

Some tech entrepreneurs say Europe is not ready.

Jonas Torland, co-founder of Norway-based 7Analytics, which develops models for predicting floods and landslides, said governments and businesses in the United States had risk managers who were more accustomed to assessing environmental hazards, while in Europe, authorities lacked readiness.

"We often see substantial expenditures with minimal data support for informed decision-making", Torland, whose models are used in the cities of Oslo, Bergen and Kristiansand, told the Thomson Reuters Foundation.

"While AI is a crucial component of these models, unfortunately, governments are not investing in or purchasing these advanced AI solutions," he said adding that he believed governments "stick to their old data providers and consultants.”

Data processing is also a challenge because these complex AI models need to run updates every hour as forecasts change.

That requires both a lot of computing power, and a lot of time - especially at more minute scales.

A 1-by-1 meter grid, which 7Analytics uses for its predictions, is 100 times more detailed than a 10-by-10 meter grid, but requires more than 100 times as long to process.

High computing power also means huge amounts of energy and water are needed, which makes AI models part of the problem because they are adding to the planet-heating emissions driving the climate emergency.

Some big technology companies, like Microsoft and Google, are exploring the use of nuclear power to run their huge data storage centers.

Other scientists stress that beyond refining their forecasting abilities, authorities need to invest in physical solutions, like developing areas where floodwater can safely be stored, and early warning systems.

They also need to minimize development in flood-prone areas, given the likelihood of more intense climate change-driven floods, and meet their commitments to limit emissions.

"It's not a question of data or technology or knowledge. It's a question of implementation, political will," Friederike Otto, a senior lecturer at Imperial College in London, said in an email response to questions.

"As long as the world burns fossil fuels, the root cause of climate change, extreme weather events will continue to intensify, killing people and destroying homes. To curb this trend, we need to replace oil, gas and coal with renewable energy."



Reddit Sues AI Giant Anthropic Over Content Use

Dario Amodei, co-founder and CEO of Anthropic. JULIEN DE ROSA / AFP
Dario Amodei, co-founder and CEO of Anthropic. JULIEN DE ROSA / AFP
TT
20

Reddit Sues AI Giant Anthropic Over Content Use

Dario Amodei, co-founder and CEO of Anthropic. JULIEN DE ROSA / AFP
Dario Amodei, co-founder and CEO of Anthropic. JULIEN DE ROSA / AFP

Social media outlet Reddit filed a lawsuit Wednesday against artificial intelligence company Anthropic, accusing the startup of illegally scraping millions of user comments to train its Claude chatbot without permission or compensation.

The lawsuit in a California state court represents the latest front in the growing battle between content providers and AI companies over the use of data to train increasingly sophisticated language models that power the generative AI revolution.

Anthropic, valued at $61.5 billion and heavily backed by Amazon, was founded in 2021 by former executives from OpenAI, the creator of ChatGPT.

The company, known for its Claude chatbot and AI models, positions itself as focused on AI safety and responsible development.

"This case is about the two faces of Anthropic: the public face that attempts to ingratiate itself into the consumer's consciousness with claims of righteousness and respect for boundaries and the law, and the private face that ignores any rules that interfere with its attempts to further line its pockets," the suit said.

According to the complaint, Anthropic has been training its models on Reddit content since at least December 2021, with CEO Dario Amodei co-authoring research papers that specifically identified high-quality content for data training.

The lawsuit alleges that despite Anthropic's public claims that it had blocked its bots from accessing Reddit, the company's automated systems continued to harvest Reddit's servers more than 100,000 times in subsequent months.

Reddit is seeking monetary damages and a court injunction to force Anthropic to comply with its user agreement terms. The company has requested a jury trial.

In an email to AFP, Anthropic said "We disagree with Reddit's claims and will defend ourselves vigorously."

Reddit has entered into licensing agreements with other AI giants including Google and OpenAI, which allow those companies to use Reddit content under terms that protect user privacy and provide compensation to the platform.

Those deals have helped lift Reddit's share price since it went public in 2024.

Reddit shares closed up more than six percent on Wednesday following news of the lawsuit.

Musicians, book authors, visual artists and news publications have sued the various AI companies that used their data without permission or payment.

AI companies generally defend their practices by claiming fair use, arguing that training AI on large datasets fundamentally changes the original content and is necessary for innovation.

Though most of these lawsuits are still in early stages, their outcomes could have a profound effect on the shape of the AI industry.