Artificial intelligence can now outperform doctors at diagnosing patients in the emergency department, according to a new study in Science.

The AI was given written notes from real emergency department records from a hospital in Boston, US, and asked to weigh in at different points during the patient’s care. At the earliest stage – triage, when a patient first arrives – the AI identified the correct diagnosis, or something closely related, in 67% of cases.

The two doctors used for comparison managed 50% and 55%. That’s a meaningful gap, especially at the moment when information is scarcest and uncertainty is highest.

This study matters because the field is moving so fast. Earlier research showed that large language models – the technology behind systems like ChatGPT – could pass medical licensing exams. Interesting, but not all that illuminating. Passing an exam is not the same as being useful on a ward.

This new study goes further. It puts AI alongside doctors across several tasks, using genuine clinical text from a real emergency department. That makes it more directly relevant to medical practice than most of what’s come before. It suggests these systems are developing into something that could genuinely help doctors think through a wide range of possible diagnoses, especially in situations where missing a serious condition is the main concern.

There are good reasons, though, not to get carried away.

The AI was working entirely from written text. It never saw the patient, never noticed how breathless or frightened they looked, never examined them, spoke to their family, weighed up the chaos of a busy department, or took any responsibility for what happened next. It was not practising emergency medicine. It was offering a written opinion based on selected information.

There’s also a gap between producing a list of possible diagnoses and actually improving patient outcomes. A longer list might help a doctor think more broadly, but it could equally generate new problems: unnecessary tests, over-treatment, extra workload, or unwarranted confidence in an answer that sounds plausible but turns out to be wrong.

And some of the benchmark cases used in studies like this may have been publicly available when the AI was trained, which doesn’t undermine the emergency department findings, but is another reason to treat headline numbers with some scepticism.

The hard question

So the question isn’t really whether AI can help doctors think through difficult cases. The harder question is how this should be tested and governed in real clinical settings like the NHS.

That question is already urgent. A Royal College of Physicians snapshot found that 16% of UK doctors were using AI tools in clinical practice every day, with another 15% doing so weekly. Doctors are already using these tools in their daily work – before hospitals and health systems have properly worked out how to assess them, train staff to use them safely, spot when they’re causing harm, or decide who is responsible when something goes wrong.

It’s tempting to say that the solution is to keep a human in the loop. But that phrase does very little work on its own. We need to know which human, in which loop, and with what authority. A doctor’s ability to override an AI suggestion is not, by itself, a safety system. Someone still has to decide which tools get used, who can change how they behave, how harms are spotted, and who is responsible when the tool quietly starts failing.

This study represents genuine progress. But it doesn’t, on its own, change how medicine should be practised. The right response is neither to prohibit these systems nor to let them quietly become part of the routine before anyone has thought it through. They should be trialled in real clinical settings, used as a form of second-opinion support rather than a substitute for clinical judgment, and measured against what actually matters to patients: care that is better, safer and faster.

Ewen Harrison receives funding from a number of grant-giving bodies including UKRI, NIHR, HDRUK, and Wellcome Leap. He is a Deputy Editor with NEJM AI.

Barring a last-minute surprise, the Holyrood election will probably return the Scottish National Party (SNP) to government for the fifth time in a row. The nationalists have been in office for so long that thousands of Scots who weren’t even born when the party entered office in 2007 are now going to the polls for the first time.

But just 23% of respondents think the Scottish government is doing a good job, according to the Scottish Election Study’s final pre-election Scoop poll in February. This is down from 44% immediately before the 2021 election. And according to the latest opinion polls, the party is on track to drop at least 10 percentage points of the 47.7% constituency vote share it recorded back then.

It was always going to be challenging for the SNP to sustain such high levels of support given the economic climate. As our public opinion tracking data demonstrates, it didn’t take long after 2021 for the public mood to sour as the second-order impacts of the COVID pandemic and longer-running economic woes sank in.

Issues such as inflation, a creaking health service, chronic housing shortages and high street stagnation dented voters’ faith in governing parties at both Westminster and Holyrood. The share of Scots who thought the country was heading in the right direction dropped from just over 40% in 2021 to under 20% by 2025.

Pro-independence voters, who had loyally voted SNP as a bloc since the 2014 referendum, began to point the finger at the party even before long-time first minister Nicola Sturgeon resigned.

The nationalists dropped in the opinion polls after enduring a year of scandal and two fraught changes of leadership, losing most of their Westminster MPs at the 2024 general election. This was their first popular vote loss to Labour since 2010.

Any other unpopular, long-in-the-tooth incumbent would be staring down the barrel of a decisive defeat. And the party’s vote share will undoubtedly decline. But two big factors will combine to buoy the SNP’s seat count at this election and likely propel the party back to power. First, the continued polarisation of the Scottish electorate on the question of independence. And second, an ever-more fragmented opposition.

Crossing the divide on independence

Although the salience of Scottish independence has declined since 2021, the SNP retains the support of two thirds of pro-independence voters. At the same time, the only other pro-independence party, the Scottish Greens, has withdrawn from all but a handful of constituencies.

Voters appear to be more willing to cross the constitutional divide than they did five years ago. But attitudes to independence continue to structure voting behaviour and views on the governments at Holyrood and Westminster. Enough voters still trust the SNP to “stand up for Scotland” within the union to stick with the party, even if they’re less enthusiastic on this occasion.

What’s more, the opposition is now even more divided. Both Labour and the Conservatives also look likely to lose support compared to peaks in recent elections. The Conservatives, the second-largest party in the previous parliament, are set to shed around half their vote from 2021. And Labour look likely to lose much of the ground they had made up by 2024 due to a faltering first two years in power at Westminster.

Combined, these parties and the SNP secured 91% of constituency votes in 2021 – this figure could drop to somewhere between 60% and 70% this time. Reform UK is competing with Labour for second place from a standing start, while the Liberal Democrats are also likely to advance. Voters hoping to unseat the SNP may agree on that, but little else.

When this is fed into the electoral system, with 73 of the parliament’s 129 seats decided by first-past-the-post constituencies, the SNP can afford to lose a sizeable chunk of support and live to fight another day. Its vote is evenly spread around the country, and the splintering opposition (not to mention changing constituency boundaries) make it difficult to unseat when there is no consensus challenger of the kind Labour looked like being two years ago.

There are, however, substantial risks here for the nationalists (assuming they remain in charge). Instead of running a “big tent” campaign resembling the pre-indyref years – and with a two-decade record to defend – the SNP has been forced to pursue a core vote strategy, hoping to maximise turnout of the left-leaning, socially progressive “Yes” base.

To this end, the Scottish Greens have done the party a big favour by retreating in most constituencies and will expect to be rewarded. A Swinney government may find itself with some very difficult budgetary choices in a tightening fiscal environment which are at odds with an expansionary manifesto.

And the so-called “scunner factor” at this election, with low turnout expected and Reform UK and Green gains virtually guaranteed, suggests that patience is running thin with mainstream parties.

While it’s unlikely to bottom out as quickly as backing for Keir Starmer’s Labour government, continued stagnation in living standards could see SNP support erode further. Then again, the nationalists’ superpower has always been to use the political weather to their advantage – and the wide-open 2029 UK general election could provide another such opportunity.

If the nationalists can hang on to power, analysts looking back in another 20 years might regard Swinney’s own “loveless landslide” as the most important SNP victory of them all.

Fraser McMillan receives funding from UKRI/ESRC as part of the Scottish Election Study.

The robot pauses at the edge of the room as an engineer checks its sensors. Then, with a soft mechanical hum, this humanoid machine begins to move. It lifts a mannequin from a bed, slowly and carefully. The engineers hold their breath.

I am in a robotics lab in Tokyo, Japan, as part of my Wellcome research fellowship. The engineers have repeated this test hundreds of times over several weeks, with mixed results.

Japan has one of the world’s oldest populations, and a strained health and care workforce. It has also long been the global leader in the development and deployment of care robots.

While other countries are exploring robotic technologies, Japan stands out for the size of its investment over several decades, and the strong link between innovation and national policy.

Government-led initiatives such as Society 5.0 and Moonshot promote a “super-smart” society in which, by 2050, robots could be integrated into everyday life. One early example is the impending trial of humanoid baggage handlers at Tokyo’s Haneda airport.

My research assesses what the introduction of robots means for care in Japan. This includes ethical and societal questions about affordability, privacy risks, data management and safety – and what the Japanese public thinks about these technologies.

In a care sector that is globally under pressure, different types of robot – from humanoid and pet-like “companions” to more straightforward mechanical aids – could prove useful. Some help lift people, reducing physical strain on care workers. Others remind a patient to take medication, support rehabilitation exercises, and monitor their vitals.

However, my research shows there is still a big gap between staged robotic demonstrations and everyday reality.

Uniquely human skills

Many of the robots I observed were tested in carefully controlled environments. Floors were cleared, lighting was adjusted, engineers stood nearby ready to step in. In some cases, the robots’ actions were partly, if not entirely, controlled from a distance.

In contrast, real care environments are busy, unpredictable and crowded. People move suddenly. Their needs change from moment to moment. Technologies that work well in labs still struggle in these settings.

A carer can notice a change in someone’s mood and adjust how they speak. They can offer comfort without being asked. These are uniquely human skills. As one family caregiver put it: “The promise of robotic care is practical, but the experience of care is emotional – that’s where the tension lies.”

Some family carers and professional careworkers welcomed the idea of robotic assistance, especially for physically demanding tasks like lifting. Others worried that too much reliance on machines could make care feel impersonal.

“To some older adults, these technologies are helpful tools,” said one careworker. “To others, they feel confusing, frustrating – a glimpse of a future they never asked for.”

Such perspectives are often missing from media narratives that focus on robot success stories. In Japan, these are shaped by government strategies and economic priorities. Innovation is never neutral, reflecting political agendas about how society should respond to ageing and labour shortages.

The challenges over care that societies face are not only technical but social, ethical and cultural. They raise questions about what care should be, how it is valued, and what kind of future we want. “Among families and caregivers, hope and hesitation sit side by side,” a technology developer told me. “Efficiency is often welcome, but not at the cost of losing the human touch.”

The future for care?

While Japan has been successful in exporting socially assistive robots such as Paro (a therapeutic robotic that resembles a baby seal) and the humanoid Pepper, China is rapidly expanding the market with more affordable, mass-produced technologies and humanoid innovation.

However, we are still a long way from the vision of care robots feeding, washing and otherwise supporting people in the way human carers do every day. Participants in my research, including technology developers, all agreed that robots should never fully replace human carers.

Technologies that assist with lifting, mobility and routine monitoring are the most likely to become widely used and ethically and socially accepted. In these areas, robots can complement human care rather than try to replace it.

Care is, at its core, a deeply human activity, not just a series of programmable tasks. It relies on relationships, trust and mutual understanding. Robots may support these processes, but they cannot replace them.

Additionally, some technologies are likely to remain expensive, available mainly to well-funded care homes or private users. This raises issues about access to good-quality care.

Care robot developments in Japan show what can be achieved through sustained investment and political support. But they also shed light on the large amount of work needed to ensure responsible research and innovation practices in this area.

The real question is not just what robots can do. It is what kind of care we want in the future – and how technology can support it without deepening inequalities, limiting access to good-quality care, and losing the power of human touch.

Giulia De Togni received funding from the Wellcome Trust. She is affiliated with the University of Edinburgh.

For many business owners, managing a crisis in silence is the default response. Companies generally prefer to deal with the fallout behind the scenes, following a simple mantra: resolve the issue and keep up the appearance that everything is “business as usual”.

However, this time, KitKat took a different approach. Instead of keeping it low-key, the brand took the incident public, transforming it into a campaign to engage the audience.

In just a few days, the incident gained worldwide traction on social media and news outlets. Audiences shifted from passive observers to active participants. This potential reputational threat ultimately became a record-breaking campaign with over 100 million views.

This is a prime example of how brands today have shifted from mere crisis management to using unexpected challenges as a way to engage audiences in real time.

Baca juga: From ‘market value’ to levelling up, the manosphere is shaped by a financial mindset

What actually happened?

The story began in late March 2026, when a truck carrying over 413,000 KitKat bars — roughly 12 tonnes — disappeared in transit between central Italy and Poland.

The timing was critical. This “unlucky” incident took place only a week before Easter and right as KitKat was debuting as an official Formula 1 partner.

Baca juga: The more commodified your job, the more likely AI can do it – lessons from online freelancing

Take their “unique” official statement, for example. It was not your standard, boring corporate press release — it was written with a playful wink. At first, the internet was convinced it was an elaborate marketing stunt or an early April Fool’s joke, but KitKat soon confirmed: this was no prank.

In their statement, KitKat verified the theft and assured the public of product safety, but they did not stop there. By adding that the culprits had exceptional taste, they shifted the tone entirely. It was no longer a PR disaster to be fixed, it was a compelling story waiting to be told.

Working with VML, KitKat launched the Stolen KitKat Tracker, a digital tool where consumers could verify their chocolate’s origins using an eight-digit pack code.

The campaign triggered a massive cultural moment. #KitKatHeist became a trending topic, sparking millions of memes and inviting other major brands to join the conversation.

Rethinking the brand response to disruption

This case highlights the evolution of real-time marketing: the ability to pivot a crisis into a cultural moment. In today’s battle for attention (attention economy), KitKat proves that holding the public’s interest is just as vital as managing the brand’s reputation.

A key factor in the campaign’s effectiveness was direct audience involvement. By using the tracker, consumers moved beyond merely reading about the heist to actively taking part in it — a classic element of gamification. It was a simple but effective approach: it gave people a tangible reason to pick up a KitKat, engage with the brand, and, most importantly, share their results.

Consumers shifted from passive buyers to active participants in a live, unfolding story. This engagement did more than just capture attention. It drove sales and refreshed the brand’s connection with its customers.

At the same time, curiosity fuels the campaign, as the public remains unsure of how much is fact and how much is fiction. This uncertainty sparks deeper discussion and sharing, transforming a simple incident into an interactive experience while refocusing attention on the product itself.

The spillover effect: The power of collaborative brand storytelling

KitKat’s response did more than just spark attention. It created a cultural vacuum that other brands rushed to fill.

Ryanair, Domino’s, and Pizza Hut joined the party with their own official statements, each offering mock condolences while subtly promoting themselves in the same breath.

The result was a snowballing effect: every brand that joined extended the story’s lifespan, reached a new audience, and reflected that visibility back onto KitKat. What began as a single brand’s crisis evolved into a shared cultural moment, where participation became the price of entry.

Brands amplifying each other is no accident. It works because the original incident provided a clear, low-stakes hook for others to latch onto. Since there was no real harm involved, the humour was accessible to everyone; the stakes were low enough to serve as an open invitation for play.

Baca juga: Money isn’t free. Here’s what to know before downloading a cashback app

Humorous crisis communication is not merely a stylistic choice; it is a conditional one. KitKat could pull it off because the incident was victimless, posed no safety risks, and involved an everyday product with no significant moral weight.

The future of the attention economy

Campaigns like this raise an important ethical question: where should brands draw the line between seizing opportunities and corporate responsibility?

KitKat’s response worked well not just because it was creative, but because of the crisis itself: it was low-stakes, harmless, and socially acceptable. Ultimately, not every disruption should become a campaign. In the world of real-time marketing, good judgment is just as important as acting quickly.

KitKat’s success proves we have moved beyond the era of one-way communication and into the era of “navigating moments”. In today’s landscape, a brand’s ability to balance instant visibility with genuine credibility is crucial to stand out.

The challenge is no longer avoiding a crisis, but knowing how to respond in a way that builds both attention and trust.

Baca juga: How this ‘dirtbag’ billionaire chose to do capitalism differently

Para penulis tidak bekerja, menjadi konsultan, memiliki saham atau menerima dana dari perusahaan atau organisasi mana pun yang akan mengambil untung dari artikel ini, dan telah mengungkapkan bahwa ia tidak memiliki afiliasi di luar afiliasi akademis yang telah disebut di atas.