Huddled round a screen showing an animation of a busy street, a group of researchers watch as a child emerges from behind a parked car and is mowed down by a truck. The simulation resets and the same scenario plays out again. And again. Each time, the pedestrian is hit, but in different weather conditions: heavy snow, dense fog or pouring rain.

The animation looks almost like a version of Grand Theft Auto (it is based on the same game engine) but this is no console title. The research team at CARISSMA, a leading research and test centre for vehicle safety in Ingolstadt, Southern Germany, is working on a significant problem for autonomous vehicles. How to make them safer in extreme weather, when their cameras and sensors are impaired by snow, fog or rain.

The CARISSMA team is part of a €9.7 million Europe-wide project called ROADVIEW— a 15-partner research consortium with members from seven countries including the UK, Germany, France and Sweden. All made possible with funding from the EU’s flagship €96 billion research and innovation programme, Horizon Europe.

Extreme weather

The 4-year project brings together some of the best expertise in Europe on autonomous vehicle safety, sensor technology and simulation modelling to achieve results that no single country could achieve on its own. The partners come from universities, research led-companies and car manufacturers, all with the overall goal of solving the extreme weather problem for self-driving vehicles.

“It is very important to bring different partners from different countries together. To bring different thinking together,” says Professor Werner Huber, Head of the CARISSMA Institute of Automated Driving.

Though companies such as Waymo and Tesla have already brought driverless taxis to public roads in the US, their roll-out in Europe has been slower because of a more cautious attitude to safety.

Experience from the US has thrown up situations where the cars have struggled in unusual conditions. In September last year, for example, torrential rain caused acute flooding in Arizona. Driverless Waymo taxis struggled in the conditions with . That led the company to suspend its operations temporarily.

“You don’t want a system that works only in ideal conditions,” says Professor Valentina Donzella, at Queen Mary University of London, whose team of engineers is part of the ROADVIEW consortium, “It’s important to be able to demonstrate safety in adverse weather or complex conditions.”

You need data, data, data and you can’t get enough from real world tests

That’s difficult because of the need to accumulate enough driving experience in extreme scenarios. “You need data, data, data and you can’t get enough [from real world tests] to find out what the car should do in a specific situation,” says Huber. That would be too dangerous on a public road.

The answer is to train the cars’ sensors and control systems in the lab or on the test track using data that simulates extreme weather conditions, something that Huber and Donzella’s teams are working on together. “The UK partners have great experience in this kind of modelling,” says Huber, “It’s not that easy to model bad weather situations into a simulation environment since you need a sensor model which 
 is really close to reality.”

In simple terms, it means taking real-vehicle motion and reproducing it in a simulation where the data from cameras, LiDAR and radar sensors is degraded to mimic what the car would see in heavy rain, snow or fog. That way, the researchers can take real-world motion and use it in extreme weather tests.

The geographical spread of the ROADVIEW partners confers an advantage, says Donzella. She points out that the nature of fog, snow and rain is not uniform across the continent and that impacts the way autonomous vehicles “see” the road.

“We realised that, depending on the country, we have different types of [fog, rain and snow],” says Donzella, “just discussing with the partners and understanding what we have to model in terms of adverse weather was really interesting.”

In the UK’s maritime climate, for example, snow tends to be wetter than in Finland or Scandinavia. “When it’s dry, it flies more [and] is more affected by wind,” says Donzella, “You have to take that into account [in the modelling].”

Critical scenarios

In Ingolstadt, Huber’s team uses the modelling expertise from Queen Mary to fool real-world cars on the institute’s test track into believing they are driving around in extreme weather conditions. That allows them to repeat critical scenarios like the child crossing the road — but in any weather.

Huber says the collaboration with the UK team has been “very fruitful” with each side bringing different expertise to the table. “UK partners have a special kind of thinking. They are critical and they are innovative and I think German partners are the same, maybe from another perspective,” he says, adding that both sides have long expertise with the automotive industry.

Donzella agrees. “It’s a matter of complementary skills, different expertise,” she says, “It’s been very, very productive.”

The project, which runs over four years, has also provided a structure for deep and long-term collaboration. “It gave us enough time to know each other, understand strengths and how to work together,” says Donzella. Members of her research team have spent time at the CARISSMA test track in Ingolstadt, Germany and researchers from that team have made the reciprocal journey to London.

That has made for a deep and effective partnership, she says. “Obviously we leverage complementary skills, but also complementary facilities, different viewpoints. Some things that can be more challenging here in the UK are maybe less challenging in Germany due to different road infrastructure and so on.”

That’s vital to building an autonomous driving system that will be safe in conditions right across the European continent. “Working together, we can bring this different cross-geographical perspective,” she says.

The ROADVIEW project comes to an end in August, but both sides intend to continue working together. “We have understood that environmental conditions are really challenging 
 and we need to consider everything together. So the complexity is very high and there are still a lot of things that we need to tackle,” says Donzella.

“I hope this isn’t the end. I hope this is the start of a long cooperation,” says Huber.

Watch to find out more: Inside the partnership helping driverless cars navigate extreme weather 

Find out ÌęÌę

Many countries are exploring how artificial intelligence might help with everything from processing taxes to determining welfare benefits. But a survey shows citizens are not as enthusiastic as their governments – and this can create real risks for democracy.

“Focusing only on short-term efficiency gains and shiny technology risks triggering public backlash and contributing to a long-term decline in democratic trust and legitimacy,” says at the Ludwig Maximilian University of Munich in Germany.

Wuttke and his colleagues asked around 1200 people in the UK to share their feelings about government actions where either a human or an AI handled the task. These hypothetical scenarios included processing tax returns, approving or rejecting welfare applications and making risk assessments about whether defendants should be eligible for bail.

Some people were only told about how AI could improve government efficiency – but others learned about both AI-related benefits and risks. These risks included difficulty in understanding how AI decisions are made, growing government dependence on AI that becomes less reversible over time and a lack of clear paths for citizens to contest and correct AI decisions.

When people became aware of AI-related risks, they reported both a significant decline in their trust in government and a feeling of losing control. For example, the percentage of participants reporting loss of democratic control in their government increased from 45 per cent to more than 81 per cent in scenarios where the government became increasingly dependent on AI for handling specific tasks.

The proportion of people demanding less AI in government also sharply increased once participants learned about the risks – rising from under 20 per cent in the baseline scenario to more than 65 per cent in any scenario where they learned about both the benefits and risks of government using AI.

Free newsletter

Sign up to The Weekly

Sign up to newsletter

Despite these results, democratic governments could make use of AI in responsible ways that retain citizens’ trust, says at the Center for Democracy & Technology in Washington DC. But she says there are few success stories of AI in government so far. Meanwhile,  there are already “quite a few failure cases” – and the stakes of these cases can be incredibly high.

For example, US state efforts to automate the processing of public benefits claims have led to tens of thousands of people being for fraud. Some of those people ended up having to file for bankruptcy or lose their homes. “Government mistakes have enormous, long-reaching impacts,” says Quay-de la Vallee.

What is artificial intelligence? It is a question that scientists have been wrestling with since the dawn of computing in the 1950s, when Alan Turing asked: “Can machines think?” Now that large language models (LLMs) like ChatGPT have been unleashed on the world, finding an answer has never been more pressing.

While their use has already become widespread, the social norms around these new AI tools are still rapidly evolving. Should students use them to write essays? Will they replace your therapist? And can they turbocharge government?

Revealed: How the UK tech secretary uses ChatGPT for policy advice

Âé¶čŽ«Ăœ has used freedom of information laws to obtain the ChatGPT records of Peter Kyle, the UK's technology secretary, in what is believed to be a world-first use of such legislation

That last question is being asked in both the US and UK. Under the new Trump administration, Elon Musk’s Department of Government Efficiency (DOGE) taskforce is eliminating federal workers and rolling out a chatbot, GSAi, to those that remain. Meanwhile, the UK prime minister, Keir Starmer, has called AI a “golden opportunity” that could help reshape the state.

Certainly, there is government work that could benefit from automation, but are LLMs the right tool for the job? Part of the problem is we still can’t agree what they actually are. This was aptly demonstrated this week, when Âé¶čŽ«Ăœ used freedom of information (FOI) laws to obtain the ChatGPT interactions of Peter Kyle, the UK’s secretary of state for science, innovation and technology. Politicians, data privacy experts and journalists – not least us – were stunned that this request was granted, given similar requests for a minister’s Google search history, say, would generally be rejected.

That the records were released suggests that the UK government sees using ChatGPT as more akin to a ministerial conversation with civil servants via email or WhatsApp, both of which are subject to FOI laws. Kyle’s interactions with ChatGPT don’t indicate any strong reliance on the AI for forming serious policy – one of his questions was about which podcasts he should appear on. Yet the fact that the FOI request was granted suggests that some in government seem to believe the AI can be conversed with like a human, which is concerning.

As Âé¶čŽ«Ăœ has extensively reported, current LLMs aren’t intelligent in any meaningful sense and are just as liable to spew convincing-sounding inaccuracies as they are to offer useful advice. What’s more, their answers will also reflect the inherent biases of the information they have ingested.

Indeed, many AI scientists are increasingly of the view that LLMs aren’t a route to the lofty goal of artificial general intelligence (AGI), capable of matching or exceeding anything a human can do – a machine that can think, as Turing would have put it. For example, in a recent survey of AI researchers, about 76 per cent of respondents said it was “unlikely” or “very unlikely” that current approaches will succeed in achieving AGI.

Instead, perhaps we need to think of these AIs in a new way. , a team of AI researchers says they “should not be viewed primarily as intelligent agents but as a new kind of cultural and social technology, allowing humans to take advantage of information other humans have accumulated”. The researchers compare LLMs to “such past technologies as writing, print, markets, bureaucracies, and representative democracies” that have transformed the way we access and process information.

Framed in this way, the answers to many questions become clearer. Can governments use LLMs to increase efficiency? Almost certainly, but only when used by people who understand their strengths and limitations. Should interactions with chatbots be subject to freedom of information laws? Possibly, but existing carve-outs designed to give ministers a “safe space” for internal deliberation should apply. And can, as Turing asked, machines think? No. Not yet.