Johansen.se

I research the intersection of artificial intelligence, natural language processing and human reasoning as the director of the Advancing Human and Machine Reasoning lab at the University of South Florida. I am also commercializing this research in an AI startup that provides a vulnerability scanner for language models.

From my vantage point, I observed significant developments in the field of AI language models in 2024, both in research and the industry.

Perhaps the most exciting of these are the capabilities of smaller language models, support for addressing AI hallucination, and frameworks for developing AI agents.

Small AIs make a splash

At the heart of commercially available generative AI products like ChatGPT are large language models, or LLMs, which are trained on vast amounts of text and produce convincing humanlike language. Their size is generally measured in parameters, which are the numerical values a model derives from its training data. The larger models like those from the major AI companies have hundreds of billions of parameters.

There is an iterative interaction between large language models and smaller language models, which seems to have accelerated in 2024.

First, organizations with the most computational resources experiment with and train increasingly larger and more powerful language models. Those yield new large language model capabilities, benchmarks, training sets and training or prompting tricks. In turn, those are used to make smaller language models – in the range of 3 billion parameters or less – which can be run on more affordable computer setups, require less energy and memory to train, and can be fine-tuned with less data.

No surprise, then, that developers have released a host of powerful smaller language models – although the definition of small keeps changing: Phi-3 and Phi-4 from Microsoft, Llama-3.2 1B and 3B, and Qwen2-VL-2B are just a few examples.

These smaller language models can be specialized for more specific tasks, such as rapidly summarizing a set of comments or fact-checking text against a specific reference. They can work with their larger cousins to produce increasingly powerful hybrid systems.

What are small language model AIs – and why would you want one?

Wider access

Increased access to highly capable language models large and small can be a mixed blessing. As there were many consequential elections around the world in 2024, the temptation for the misuse of language models was high.

Language models can give malicious users the ability to generate social media posts and deceptively influence public opinion. There was a great deal of concern about this threat in 2024, given that it was an election year in many countries.

And indeed, a robocall faking President Joe Biden’s voice asked New Hampshire Democratic primary voters to stay home. OpenAI had to intervene to disrupt over 20 operations and deceptive networks that tried to use its models for deceptive campaigns. Fake videos and memes were created and shared with the help of AI tools.

Despite the anxiety surrounding AI disinformation, it is not yet clear what effect these efforts actually had on public opinion and the U.S. election. Nevertheless, U.S. states passed a large amount of legislation in 2024 governing the use of AI in elections and campaigns.

Misbehaving bots

Google started including AI overviews in its search results, yielding some results that were hilariously and obviously wrong – unless you enjoy glue in your pizza. However, other results may have been dangerously wrong, such as when it suggested mixing bleach and vinegar to clean your clothes.

Large language models, as they are most commonly implemented, are prone to hallucinations. This means that they can state things that are false or misleading, often with confident language. Even though I and others continually beat the drum about this, 2024 still saw many organizations learning about the dangers of AI hallucination the hard way.

Despite significant testing, a chatbot playing the role of a Catholic priest advocated for baptism via Gatorade. A chatbot advising on New York City laws and regulations incorrectly said it was “legal for an employer to fire a worker who complains about sexual harassment, doesn’t disclose a pregnancy or refuses to cut their dreadlocks.” And OpenAI’s speech-capable model forgot whose turn it was to speak and responded to a human in her own voice.

Fortunately, 2024 also saw new ways to mitigate and live with AI hallucinations. Companies and researchers are developing tools for making sure AI systems follow given rules pre-deployment, as well as environments to evaluate them. So-called guardrail frameworks inspect large language model inputs and outputs in real time, albeit often by using another layer of large language models.

And the conversation on AI regulation accelerated, causing the big players in the large language model space to update their policies on responsibly scaling and harnessing AI.

But although researchers are continually finding ways to reduce hallucinations, in 2024, research convincingly showed that AI hallucinations are always going to exist in some form. It may be a fundamental feature of what happens when an entity has finite computational and information resources. After all, even human beings are known to confidently misremember and state falsehoods from time to time.

The rise of agents

Large language models, particularly those powered by variants of the transformer architecture, are still driving the most significant advances in AI. For example, developers are using large language models to not only create chatbots, but to serve as the basis of AI agents. The term “agentic AI” shot to prominence in 2024, with some pundits even calling it the third wave of AI.

To understand what an AI agent is, think of a chatbot expanded in two ways: First, give it access to tools that provide the ability to take actions. This might be the ability to query an external search engine, book a flight or use a calculator. Second, give it increased autonomy, or the ability to make more decisions on its own.

For example, a travel AI chatbot might be able to perform a search of flights based on what information you give it, but a tool-equipped travel agent might plan out an entire trip itinerary, including finding events, booking reservations and adding them to your calendar.

AI agents can perform multiple steps of a task on their own.

In 2024, new frameworks for developing AI agents emerged. Just to name a few, LangGraph, CrewAI, PhiData and AutoGen/Magentic-One were released or improved in 2024.

Companies are just beginning to adopt AI agents. Frameworks for developing AI agents are new and rapidly evolving. Furthermore, security, privacy and hallucination risks are still a concern.

But global market analysts forecast this to change: 82% of organizations surveyed plan to use agents within 1-3 years, and 25% of all companies currently using generative AI are likely to adopt AI agents in 2025. Läs mer…

The “choking game” has potentially deadly consequences, as players are challenged to temporarily strangle themselves by restricting oxygen to the brain. It sounds terrifying, but rough estimates suggest that about 10% of U.S. teenagers may have played this type of game at least once.

There’s more, unfortunately: The Skullbreaker Challenge, the Tide Pod Challenge and Car Surfing are but a few of the deadly games popularized through social media, particularly on Snapchat, Instagram, TikTok, YouTube and X – formerly Twitter. Many of these games go back more than a generation, and some are resurging.

The consequences of these so-called games can be deadly. Skullbreaker Challenge, for example, involves two people kicking the legs out from under a third person, causing them to fall and potentially suffer lasting injuries. Swallowing detergent pods can result in choking and serious illness. A fall from car surfing can lead to severe head trauma.

Coming up with an exact number of adolescent deaths from these activities is difficult. Data is lacking, partly because public health databases do not track these activities well – some deaths may be misclassified as suicides – and partly because much of the existing research is dated.

A 2008 report from the Centers for Disease Control and Prevention found that 82 U.S. children over a 12-year period died after playing the Choking Game. About 87% of the participants were male, most were alone, and their average age was just over 13. Obviously, new, updated research is needed to determine the severity of the problem.

Arriani played the ‘Blackout Challenge.’

Peer pressure and the developing brain

We are a professor of educational neuroscience and a Ph.D. student in educational psychology. Both of us study how children regulate their behaviors and emotions, why teenagers are particularly vulnerable to dangerous games, and how social media amplifies their risks.

Risk-taking is a necessary part of human development, and parents, peers, schools and the broader community play an integral role in guiding and moderating risk-taking. Children are drawn to, and often encouraged to engage in, activities with a degree of social or physical risk, like riding a bike, asking someone for a date or learning how to drive.

Those are healthy risks. They let children explore boundaries and develop risk-management skills. One of those skills is scaffolding. An example of scaffolding is an adult helping a child climb a tree by initially guiding them, then gradually stepping back as the child gains confidence and climbs independently.

Information-gathering is another skill, like asking if swallowing a spoonful of cinnamon is dangerous. A third skill is taking appropriate safety measures – such as surfing with friends rather than going by yourself, or wearing a helmet and having someone nearby when skateboarding.

The perfect storm

During adolescence, the brain is growing and developing in ways that affect maturity, particularly within the circuits responsible for decision-making and emotional regulation. At the same time, hormonal changes increase the drive for reward and social feedback.

All of these biological events are happening as teenagers deal with increasingly complex social relationships while simultaneously trying to gain greater autonomy. The desire for social validation, to impress peers or to attract a potential romantic interest, coupled with less adult supervision, increases the likelihood of participating in risky behaviors. An adolescent might participate in these antics to impress someone they have a crush on, or fit in with others.

That’s why the combination of teenagers and social media can be a perfect storm – and the ideal environment for the proliferation of these dangerous activities.

Climbing a tree could be a healthy and safe way to learn about taking risks.
Johner Images via Getty Images

Social media shapes brain circuits

Social media platforms are driven by algorithms engineered to promote engagement. So they feed you what evokes a strong emotional reaction, and they seem to prioritize sensationalism over safety.

Because teens strongly react to emotional content, they’re more likely to view, like and share videos of these dangerous activities. The problem has become worse as young people spend more time on social media – by some estimates, about five hours a day.

This may be why mood disorders among young people have risen sharply since 2012, about the time when social media became widespread. These mood disorders, like depression and conduct issues, more than double the likelihood of playing dangerous games. It becomes a vicious cycle.

Rather than parents or real-life friends, TikTok, YouTube and other apps and websites are shaping a child’s brain circuits related to risk management. Social media is replacing what was once the community’s role in guiding risk-taking behavior.

Protecting teens while encouraging healthy risk-taking

Monitoring what teens watch on social media is extraordinarily difficult, and adults often are ill-equipped to help. But there are some things parents can do. Unexplained marks on the neck, bloodshot eyes or frequent headaches may indicate involvement in the choking game. Some social media sites, such as YouTube, are sensitive to community feedback and will take down a video that is flagged as dangerous.

As parents keep an eye out for unhealthy risks, they should encourage their children to take healthy ones, such as joining a new social group or participating in outdoor activities. These healthy risks help children learn from mistakes, build resilience and improve risk-management skills. The more they can assess and manage potential dangers, the less likely they will engage in truly unhealthy behaviors.

But many parents have increasingly adopted another route. They shield their children from the healthy challenges the real world presents to them. When that happens, children tend to underestimate more dangerous risks, and they may be more likely to try them.

This issue is systemic, involving schools, government and technology companies alike, each bearing a share of responsibility. However, the dynamic between parents and children also plays a pivotal role. Rather than issuing a unilateral “no” to risk-taking, it’s crucial for parents to engage actively in their children’s healthy risk-taking from an early age.

This helps build a foundation where trust is not assumed but earned, enabling children to feel secure in discussing their experiences and challenges in the digital world, including dangerous activities both online and offline. Such mutual engagement can support the development of a child’s healthy risk assessment skills, providing a robust basis for tackling problems together. Läs mer…

Threats and harassment are pushing some politicians out of office, scaring off some would-be candidates and even compelling some elected officials to change their vote.

Those are some of the conclusions of a new study I led on political violence in Southern California.

Rising threats against public officials is a national problem.

Between 2013 and 2016, there were, on average, 38 federal charges involving threats to public officials per year, according to the National Counterterrorism Innovation, Technology and Education Center, a research center. That average sharply increased between 2017 and 2022, when an average of 62 federal charges were brought annually for threats to public officials.

When elected officials worry for their safety, it has implications for all Americans. Democracy suffers when people are governed by fear.

‘Respectful discourse has been lost’

I am the founder and director of the Violence, Inequality and Power Lab, or VIP Lab, housed at the University of San Diego’s Kroc Institute for Peace and Justice. Over the past two years, the VIP Lab has been collecting data to understand the frequency and severity of threats against local elected officials in Southern California.

Our research focused on California’s three southernmost counties – San Diego, Riverside and Imperial. Together, these counties have just under 6 million residents, or roughly 15% of California’s population.

To capture as complete a picture as possible, we did a survey and interviews, reviewed news coverage and social media accounts, and scoured literature nationwide.

The first year, we focused only on San Diego County, surveying 330 mayors, city councilors, county board of supervisor members and school board and community college board members. Over 25% of survey recipients responded. Of them, 75% reported being threatened or harassed at least once in the past five years. Roughly half said the abuse occurred at least monthly.

Respondents had found their name shared on the dark web and seen cars drive past their homes in an intimidating manner. They’d been followed after public meetings and blocked from leaving. In some cases, their families were harassed.

“As a parent, [I] feel vulnerable,” one city council member said, adding that he’s become “very guarded with [my] kid in public.”

Topics that were most likely to prompt threats and harassment included COVID-19, gun control, school curricula and LGBTQ+ rights.

“Since the pandemic, people have been mobilized into different silos or groups of people,” said a school board member interviewed in 2023. “[R]espectful discourse has been lost in all of this.”

A resident and elected official in a heated debate after a city council meeting, South Gate, Calif., 2003.
Richard Hartog/Los Angeles Times via Getty Images

In year two, we sent surveys to 785 elected officials in all three counties. Two-thirds of respondents reported having been threatened or harassed at least once in the previous five years. Roughly the same number said verbal attacks had become a routine part of public service.

These attacks come from the public, they told us, and from other elected officials. Officials have been accused of corruption, called idiots and told they should die. School board members face allegations that they “don’t care about kids.”

The threats “are verbal, at council meetings, outside of meetings, during breaks,” said one interviewee serving on a city council. “I’ve been harassed by city council members, staff members, the city manager and the city attorney.”

A troubling trend

In simple terms, our research suggests that at least two of every three people who serve in public office in Southern California will be threatened, intimidated or harassed during their tenure.

Survey results suggest the average female elected official who experiences abuse is threatened or harassed at least six times as often as her male peers. Men reported being on the receiving end of abuse about once a year, while women suffer abuse almost monthly.

The attacks against women are more likely to be personalized – referring to their looks or their family members – and have a sexual nature.

It was “slanderous stuff,” one school board member told us of abusive text messages that started in 2022 after many years of service. “Language of being evil … of not being a Christian woman.”

Her husband was also followed by a car, and her home was circled by the same vehicle. No one else on her board reported similar abuse.

We heard many accounts like this from female elected officials in Southern California. One city councilwoman filed two police reports against men who threatened, harassed and stalked her. A second was threatened throughout her campaign and time in office, including by a man who used a racial slur and threatened to “take care of” her with his AK-47.

Even so, our most recent survey revealed that male elected officials are most concerned about political violence. Sixty-four percent reported that things had become worse during their time in office, compared with 50% of women.

Counterintuitively, white, male, rural and conservative respondents all reported that threats and harassment had gotten worse more often than their nonwhite, female, urban and liberal counterparts – even though nonwhite, female, urban and liberal respondents reported more threats and harassment overall.

This finding may reflect a meaningful shift in how threats are used in politics. We believe that those responsible for abuse previously targeted the most vulnerable elected officials – namely women and other underrepresented groups.

But as it becomes more common to use threats and harassment as a means to influence decision-making, everyone is a target.

Most of the abuse we documented is, thankfully, not physical. But “hostile, aggressive or violent acts motivated by political objectives or a desire to directly or indirectly affect political change or change in governance” is, by definition, political violence.

And our research shows that this constant, low-level abuse is taking its toll on people and communities.

Fear-based governing

Our study results mirror findings from other research on growing political violence in the U.S.

The number of threats targeting members of Congress went up 88% between 2018 and 2021, from 5,206 in 2018 to 9,625 in 2021.

Meanwhile, a 2023 study on state legislators by the nonprofit Brennan Center for Justice found that 89% had been threatened, harassed or insulted at some point over the previous three years. That means roughly 6,000 of the approximately 7,000 state legislators in the U.S. have been abused or intimidated since 2020.

The Capitol insurrection of Jan. 6, 2021, demonstrated for many Americans the threat of political violence.
Brent Stirton/Getty Images

Most Americans don’t need these data points: Three-quarters of Americans already believe political violence is a problem, according to the States United Democracy Center.

Constituents have a right, even an imperative, to make their opinions known to the individuals they elect. Accountability and representation are essential to democracy. But there is a line between expressing disagreement and using intimidation or violence to influence policy decisions. And the latter can have some distinctly undemocratic outcomes.

Six percent of the elected officials we interviewed said they had actually changed their vote on a specific issue due to the climate of fear. And 43% of our survey respondents said that threats and harassment have caused them to consider leaving their post.

“I don’t think it’s fair to have to fight so hard,” said one relatively new school board member. “I’m mad at myself for letting the bullies win.”

The climate of fear is also keeping people from serving. Nationwide, 69% of mayors surveyed by the Mayors Innovation Project said they knew someone who had decided not to run for office due to threats or fear of violence.

When fear – rather than the needs of community – becomes a driving force in politics, democracy loses. That’s rule by the powerful, not rule by the people. Läs mer…

The work of crafting reparations at the municipal level is fierce.

Detroiters know. In November 2021, residents voted to create a reparations committee that would make recommendations for housing and economic development programs to address historical discrimination against Black residents.

Three years have passed – and Detroiters recently learned the report is delayed. Some folks are wondering: What is taking so long?

We are a team of University of Michigan-based scholars of political science and sociology specializing in public opinion and attitudes toward reparations. Our research provides important context for understanding the challenges Detroit’s reparations committee faces.

Detroit’s missed deadline

The 13-member task force wasn’t officially announced until February 2023. The two-year time lapse resulted from the COVID-19 pandemic and the unexpected passing in July 2022 of former City Council Member JoAnn Watson – a pivotal advocate for reparations.

Its membership included three co-chairs appointed by City Council President Mary Sheffield and nine at-large members selected by the council itself.

The task force held its first meeting in April 2023. At that time, it was given 18 months to produce a report outlining harms experienced by Black Detroiters and recommendations for reparations.

As the October 2024 deadline approached, however, many residents voiced concerns that the task force was not progressing quickly enough or being as inclusive as expected. When the deadline extension was announced, task force co-chair Keith Williams issued his own report. This unilateral decision prompted the task force to release a statement letting Detroiters know that Williams’ report is not representative of the final report that is still in development. The official report is now due in March 2025.

It’s hard all over

The challenges faced by Detroit’s reparations task force are not unique to Detroit.

In the past five years, 19 reparations initiatives have been introduced and passed in U.S. cities such as San Francisco, Evanston, Illinois, and Providence, Rhode Island.

Every locality tasked with developing reparations recommendations has faced similar challenges. These include providing evidence of past and current harm, developing plans to rectify those harms, defining who is eligible to receive reparations benefits, and ensuring transparency in developing and implementing these programs.

This is challenging work, requiring navigating bureaucratic systems and political constraints while dealing with public opinion.

Nonetheless, in the absence of federal action, the country is experiencing a rise in reparations efforts enacted into local law, so figuring out what works is worth investigating.

Why achieving reparations is difficult

Public policymaking requires a fine balance between what is administratively sustainable and what is politically acceptable.

An administratively sustainable reparations policy requires a secured source of long-term funding, trained staff and established rules and procedures for efficient implementation.

A politically acceptable reparations policy requires support and acceptance by relevant political figures and the general public.

These two elements are frequently in tension.

For example, a reparations committee could suggest reallocating money from a city’s parks and recreation fund. That would be sustainable, but it is unlikely to be politically successful due to strong public opposition to interference with parks and recreation funding.

The Detroit committee has not yet developed reparations recommendations. It has requested a harms report from a network of University of Michigan scholars and an ethnohistory of Detroit’s Black community from scholars at Columbia University. These reports are now complete and will be used to assist the task force in crafting its reparations recommendations to city council.

Another positive sign: Our research suggests it can expect goodwill from the public.

Support for reparations in Detroit, including among white residents, is higher than the national average, increasing the likelihood of passing reparations initiatives and minimizing future legal pushback from non-Black residents.

And yet, it is still true that if the committee fails to maintain a balance between sustainability and political success, it risks losing this crucial public support. So where can it turn for inspiration?

Learning From Evanston

So far, Evanston, Illinois, is the only city to succeed in providing reparations to its African American population. The city’s US$10 million program is funded through a tax on the sale of recreational cannabis.

Robin Rue Simmons
AP Photo/Teresa Crawford

The effort began in 2002 when Evanston City Council member Lionel Jean-Baptiste proposed a reparations plan. The plan didn’t gain momentum until 17 years later when it was reintroduced by Councilwoman Robin Rue Simmons and passed.

Just two years later, reparations were issued to a first group of Black residents who lived in Evanston between 1919 and 1969, or who are direct descendants of people who lived in the city then. Eligible residents received a $25,000 grant to be used for housing, including repairs on existing homes. As of January 2024, 117 qualified residents had received a total of $2,953,596. There is a waitlist for several hundred people who will receive money as it comes available.

Evanston, Ill., is the first city in the nation to grant reparations to residents.
Scott Olson/Getty Images

Critics of the Evanston program point out that the program helps only a handful of people in a city with more than 12,000 African American residents. Scholars Monique Newton and Matthew Nelsen argue that it did not adequately meet the needs of the Black people for whom it was intended to repair harm. The researchers found that many of the city’s Black residents felt the program’s focus on addressing homeownership left many residents behind.

For example, Black renters in Evanston were less likely than Black homeowners to claim reparations funds. Newton and Nelsen noted that the initiative’s quick work, “while well-intentioned, contributed to a policy that left many questions about eligibility, funding, and implementation unanswered and threatened the very viability of the policy.”

Many Black residents in Evanston wanted to see more than the enactment of a reparative policy. They wanted a reparative policymaking process, during which policymakers invested time into listening to local Black residents and who designed a policy that showed an attentiveness to their preferences, concerns and experiences in a way that most American policy does not.

Implications for Detroit

In our opinion, Detroit can build a reparations program that’s fair and effective. This will mean taking the time to strike a balance between what’s practical and what people will support. The task force’s final report will need to consider funding feasibility, set up clear and obtainable steps for how the plan will work, and be transparent so the public can stay informed and confident in the process.

The task force is taking a positive step by seeking more public engagement through surveys and hosting monthly two-hour, in-person meetings to engage the public.

Despite criticisms of the task force’s efforts so far, we feel going slow allows for more time to gather input from a broader range of Black voices, ensuring that diverse concerns are fully considered.

By carefully creating its own course, Detroit’s reparations task force can not only right historical wrongs but also set a positive example for other cities looking to do the same. Läs mer…

Pets are increasingly viewed as integral family members. This often means owners want to offer their pets the same experiences as our human friends and family. But the holidays can bring added stress, worry and uncertainty for some pets, including noise, visitors, altered routines and potential exposure to dangerous foods, toys and decorations.

Vets report that Christmas is one of the busiest times of the year – and a festive vet visit is the gift everyone wants to avoid. So, it’s a good idea to do some careful management and preparation to keep pets – and the people around them – safe during the seasonal break.

Giving gifts and seeing the recipient open them is a simple pleasure. Dogs, in particular, seem to relish the opportunity to tear paper and boxes apart – and the chance to rip into something with permission can be an enriching and stimulating activity for them.

Avoid using glitter or fancy decorations for pet gifts. A cardboard box wrapped in newspaper, stuffed with scrunched up paper – perhaps with tasty treats hidden inside – is a far more appreciated gift for an active dog than a carefully wrapped toy (that might only remain intact for minutes). This homemade gift can keep dogs occupied far longer too – ideal when you might be busy entertaining, preparing Christmas dinner or just winding down.

Festive food

Treats, chews and festive foods abound at Christmas. Unfortunately, overindulging in high fat, high calorie snacks is as bad for pets as it is for humans. Avoid the temptation to feed human foods to your pet. Sudden dietary changes can result in digestive problems in our pets – not something we want to deal with during the party season.

Be aware of foods that are potentially toxic for animals. Grapes, raisins, sultanas and chocolate are all poisonous to dogs, for example. And onions are harmful for dogs and cats if ingested.

Take care with seasonally themed pet treats often available at Christmas. Brightly coloured rawhide chews might look like fun choices for dogs, but they have little nutritional value, can cause digestive upsets and, in extreme cases, swallowing large pieces can lead to digestive blockages.

Pay attention to labelling of pet treats, too. Ingredient lists may not be accurate and, sadly, some manufacturers are still lagging behind on animal welfare standards, so look for place of origin on labels and ask suppliers if you are unsure. Treats and chews can also be contaminated with microorganisms such as salmonella, which can make both pets and people ill, so source such items carefully and store them appropriately before feeding.

Natural chews such as furry rabbit ears are increasingly popular for pet dogs. Contrary to popular opinion, though, they are not an effective way to manage parasites. But they can be a useful source of animal-derived dietary fibre to help support digestive health in dogs and cats.

To help keep pets healthy and happy over the holidays, avoid overindulging them with edible gifts and make any dietary changes slowly.

Christmas costumes

Novelty pet outfits might seem like adorably cute gifts, but, for many pets, being forced to wear novelty clothing can be uncomfortable and cause distress.

Sometimes functional pet clothing is useful and beneficial for warmth and protection in winter but, even then, it is important that pets are carefully trained to wear these items through positive association.

Read more:
Five reasons your dog might really benefit from canine clothing

The best advice is to keep the ugly Christmas jumper wearing for humans, unless your pet is already happy and familiar with wearing them.

Your local pet shop will probably have a wide range of Christmas-themed toys. Although these toys should be designed for safety, pets should always be carefully supervised while playing. Many dogs have an unfortunate tendency to swallow non-edible items that need surgery to remove, including festive decorations. Throw away any damaged toys that might cause injury.

Cats often enjoy playing with simple toys such as elastic hairbands attached to string and moved around. You might find toys that contain attractive scents such as cat nip, but ensure the scent is species appropriate and always keep toy use under supervision.

Read more:
How to train your dog in basic ’life’ skills – and why it’s important

Free gifts

Often the simplest gifts are the most appreciated. For our pets, this is no different. Holidays can mean more time with pets. Consider taking the opportunity to have some fun with your pet by giving them the gifts of extra time and attention by doing the following:

Positive training to enhance relationships with pets as well as mentally and physically stimulating them
Playing with dogs can help promote bonding and more relaxed behaviour – useful at a busy time of year
Horses will benefit from exercise and attention in daylight hours – often a novelty for horse and rider during the dark winter months. Even chatting to your horse can enhance your relationship – a cheap and easy gift to give
Small pets, such as guinea pigs and chinchillas, might enjoy being handled and being involved in family events if they are relaxed with humans. If not, a quiet, cosy refuge is a better idea
Rabbits often dislike being handled or lifted, so give them the gift of space this Christmas. They might also appreciate some good quality hay, which is also excellent dietary fibre, essential for their digestive health.

Do always be careful to monitor your pet carefully for signs of distress or worry. Sometimes routine changes and too much attention are not appreciated. Cats are often unsure of new people in their home or significant routine changes, so a perfect present could be a safe place to hide during festivities.

Whatever gift you do decide to give to your pet, a little thought and care can make sure it is safe. Then, at the very least, the holidays will mean an opportunity for rest and relaxation for you and your pet together without unwanted vet visits. Läs mer…

Two influential tech CEOs have sparked a debate about AI’s effect on jobs. Klarna CEO Sebastian Siemiatkowski announced the buy-now, pay-later firm stopped hiring a year ago, claiming that AI can do most of the jobs that humans do. And the chief executive of AI firm Nvidia, Jensen Huang, has argued that workers are not going to lose their jobs to AI – but they could lose it to somebody who uses AI.

These opposing views capture a crucial tension as generative AI rapidly transforms workplaces. But research suggests the reality is more complex than either complete job displacement or simple augmentation.

As of spring 2024, 66% of employment in the US was exposed to a high or moderate level of generative AI impact, with the remaining 34% of jobs still affected in secondary tasks. Unlike previous technological revolutions, generative AI is being adopted at an unprecedented pace, outstripping the adoption rates of personal computers and the internet.

What makes this transformation particularly significant is that generative AI is disrupting a different array of “cognitive” and “non-routine” tasks, especially in middle- to higher-paid professions. This marks a departure from previous waves of automation that primarily affected routine manual labour. With the constantly evolving generative AI landscape, this trend is likely to continue and strengthen.

Research provides insights into how AI is actually affecting worker productivity and job security. One of the first comprehensive studies examined the effect of AI-powered conversational assistants on customer support agents. The findings challenge both the utopian and dystopian visions of AI’s impact.

The study found that access to AI tools increased worker productivity by an average of 14%. However, the benefits weren’t distributed equally. Novice and low-skilled workers saw the most dramatic improvements, with productivity gains of up to 35%. The AI effectively helped newer workers quickly adopt the techniques of high-performing agents.

Surprisingly, highly skilled workers saw minimal productivity gains from the AI tools. This suggests that rather than replacing human expertise, AI might help to make things more equal by elevating the capabilities of less-experienced workers.

Klarna’s experiment

Against this backdrop, Klarna’s approach of ceasing hiring and allowing natural workforce attrition represents a bold experiment. Siemiatkowski’s claim that AI can replace most human jobs aligns with concerns about widespread job displacement.

However, the company’s strategy of increasing salaries for remaining workers suggests a more nuanced reality: even as AI automates certain tasks, human expertise becomes more valuable, not less. This aligns with research showing that organisations need workers who can use AI tools effectively.

The boss of Klarna has said the fintech giant hasn’t had to recruit in more than a year thanks to AI.
rarrarorro/Shutterstock

The perspective of Nvidia CEO Huang – that people will lose jobs to AI users rather than AI itself – finds support in emerging workplace trends. A KPMG survey of US executives found that while administrative jobs face significant risk from AI, the effects vary widely across sectors. In manufacturing, for instance, 20% of respondents expect benefits from AI, while 24% anticipate negative effects.

This suggests that rather than wholesale replacement of workers, it’s likely to be a transformation of roles. The key factor in job security may not be whether a role can be automated, but whether workers and organisations can effectively integrate AI tools into their workflows.

Crucially, researchers studying ChatGPT adoption in Denmark found that half of the workers surveyed have used the generative AI tool. Younger, less-experienced, higher-achieving, and especially male workers were leading adoption. The study suggests that while workers see substantial productivity potential in ChatGPT, employer restrictions and a need for training are hindering its full adoption; it does not appear that workers fear job redundancy as a reason for avoiding the technology.

The evidence points to several strategies for workers and organisations navigating this transition.

1. The importance of lifelong learning

Workers need to continuously update their skills, focusing particularly on capabilities that complement AI, such as critical thinking and complex problem-solving.

2. The value of AI literacy

Understanding how to use AI tools effectively is becoming as crucial as traditional job skills.

3. The need for organisations to adapt

Companies must invest in employee training and development while implementing clear guidelines for AI use.

The rapid adoption of generative AI means these changes are happening faster than previous technological transformations. Workers and organisations that adapt quickly will probably have significant advantages over those that delay.
x
The contrasting views of Siemiatkowski and Huang represent different possibilities in our AI-augmented future. The evidence suggests neither complete displacement of human workers nor business-as-usual is likely. Instead, we’re entering a period where the ability to work alongside AI will increasingly determine career success.

Early evidence suggesting that AI can help level the playing field for less experienced workers is encouraging. But perhaps the most important lesson from both CEOs’ predictions is this: while we can’t control how AI will transform industries, we can control how well we adapt to it.

In this new landscape, the key to job security isn’t fighting AI or passively accepting it – it’s actively choosing to become the person using AI rather than the person being replaced by someone who does. Läs mer…

For a long time, dieting and exercise were the only realistic options for many people who wanted to lose weight, but recent pharmaceutical advances have led to the development of weight loss drugs. These are based on natural hormones from the intestine that help control food intake, such as GLP and GIP.

GLP-1-based drugs such as semaglutide (Wegovy) and tirzepatide (Mounjaro) work by helping people to feel less hungry. This results in them eating less – leading to weight loss.

Studies show that these drugs are very effective in helping people lose weight. In clinical trials of people with obesity, these drugs lead to a weight loss of up to 20% body weight in some instances.

But it’s important to note that not all the weight lost is fat. Research shows that up to one-third of this weight loss is so-called “non-fat mass” – this includes muscle and bone mass. This also happens when someone goes on a diet, and after weight loss surgery.

Muscle and bone play very important roles in our health. Muscle is really important for a number of reasons including that it helps us control our blood sugar. Blood sugar control isn’t as good in people who have lower levels of muscle mass.

High blood sugar levels are also linked to health conditions such as type 2 diabetes – where having high blood sugar levels can lead to blindness, nerve damage, foot ulcers and infections, and circulation problems such as heart attacks and strokes.

We need our bones to be strong so that we can carry out our everyday activities. Losing bone mass can increase our risk of fractures.

Researchers aren’t completely sure why people lose fat-free mass during weight loss – though there are a couple of theories.

It’s thought that during weight loss, muscle proteins are broken down faster than they can be built. And, because there’s less stress on the bones due to the weight that has been lost, this might affect normal bone turnover – the process where old bone is removed and new bone is formed leading to less bone mass being manufactured than before weight loss.

Exercise can help mitigate some of these effects.
shurkin_son/ Shutterstock

Because GLP-1 drugs are so new, we don’t yet know the longer-term effects of weight loss achieved by using them. So, we can’t be completely sure how much non-fat mass someone will lose while using these drugs or why it happens.

It’s hard to say whether the loss of non-fat mass could cause problems in the longer term or if this would outweigh the many benefits that are associated with these drugs.

Maintaining muscle and bone

There are many things you can do while taking GLP-1 drugs for weight loss to maintain your muscle and bone mass.

Research tells us that eating enough protein and staying physically active can be helpful in reducing the amount of non-fat mass that is lost when losing weight. One of the best types of exercise is doing resistance training or weight training. This will help to preserve muscle mass, and protein will help us maintain and build muscle.

If you find it difficult to find the time to squeeze in a workout, there are plenty of simple things you can do during your day to help maintain your muscle mass during weight loss. Using the stairs rather than taking the lift, carrying the shopping in from the car in fewer trips and walking at a slightly quicker pace are all really great ways of incorporating activity into your day.

GLP-1 drugs such as Wegovy and Mounjaro are exciting new tools for weight loss. But evidence suggests that to get the best long-term result from these drugs, it’s important to combine them with healthy changes to diet and physical activity. Läs mer…

When Elon Musk took over Twitter and changed its name to X, many users vowed to move to another platform. First was talk of a shift to Mastodon that never seemed to catch on. Then Meta tried to make Threads appealing by linking the app to Instagram – but this hasn’t had much cut-through either.

Now it’s Bluesky’s turn. In the weeks since the US election, the platform has grown rapidly as users flee X.

This is partly because Musk’s involvement in Donald Trump’s election campaign clearly showed where his political allegiances lay. But he has also fundamentally changed how people see X as a platform, particularly as a place to get good-quality journalism.

Many of those flocking to Bluesky have been publishers, news organisations and journalists, sending a message that they no longer see X as a valuable space to post their reporting. A group of UK journalists signed an open letter calling X “no longer a useful tool for objective reporting”.

Right now, Bluesky still remains very small compared to X. Getting official and widely agreed-upon user numbers is difficult, but estimates put X’s worldwide userbase at around 600 million, with small declines in recent months.

In comparison, Bluesky reported 13 million users in October, with increases following the US election. Estimates vary wildly, however, if you choose different metrics like daily active or monthly active users.

Our 2024 Digital News Report data at the Reuters Institute for the Study of Journalism (from surveys fielded in January) puts the proportion of adults using X for any purpose at 25% in both the US and UK. Meanwhile, 15% of people in the US and 14% of people in the UK report using X for news. Our figures are similar to those of Pew Research Center and Ofcom.

As for Bluesky, usage barely registered in our 2024 data. Usage for any purpose was 2% in the US and 1% in the UK.

Facebook and YouTube are still the most widely used platforms in the UK and US, followed by Instagram.

But what is perhaps more important right now is looking at the specific organisations and individuals that are leaving X. NPR was one of the first news brands to stop posting, and the Guardian followed. The European Federation of Journalists, which represents nearly 300,000 journalists, has said it will stop posting on X from January, when Trump takes office. And some celebrities are also leaving.

Journalism on Twitter

Relative to its size, Twitter/X has always had an outsized impact on public discourse. This is largely due to the types of people who were attracted to the platform and became heavy users: journalists, politicians, and other public figures. “Power users” on the platform drove public narratives in many ways. Twitter was where cultural and political debates happened. News broke on Twitter.

Our research at the Reuters Institute from 2021 and 2023 showed this. People saw the platform as a good place to see breaking news, follow news publishers, and generally keep up with goings on. While Facebook was seen as a place to connect with friends and family, and YouTube a place to find entertainment, Twitter’s brand was as a destination for up-to-date information.

Musk’s changes to X, from complicating the “blue tick” verification system, to removing news headlines from links, to deprioritising news and links, have negatively affected the news industry, making the platform less appealing to publishers.

Elon Musk has made several changes to X that have made it more difficult for journalists to share information.
Frederic Legrand/Shutterstock

As news brands and journalists leave, in their place Musk has appeared to boost the voices of right-leaning users, while abandoning many of Twitter’s content moderation policies and reframing X as a right-wing social media platform.

Bluesky, on the other hand, has demonstrated appeal to the news industry, with reports of better referral traffic, a chronological timeline that suits breaking news, the ability to customise feeds and a general creative spark that some say reminds them of early Twitter.

For those remaining on X, the experience for many has been a steady increase in misinformation, hate speech and general toxicity, fuelled by Musk’s changes to the platform. With many reliable news publishers leaving, there is the risk that less reliable sources are further amplified.

The new ‘it’ place

It is possible, because of the migration of certain types of users, that the popularity of Bluesky is overstated. If journalists, academics and other professionals see all their friends and colleagues popping up on Bluesky, it may give the impression that a wider movement is underway – but this could be a mistake.

On the other hand, it may be precisely who is moving to Bluesky that matters. What the shift among social media power users (journalists, politicians, public figures) may do is encourage others to join.

If the buzz maintains its current level, more people may sign up to be part of it. Once there, they may see Bluesky’s “anti-toxic” features, including the ability to hide replies and filter notifications, and come to like the user experience, as many journalists have.

The key to whether it can really “replace” Twitter as a driver of news will be in how many people ultimately move. What’s always required to make a platform successful is a critical mass – enough people being there to make it feel useful and interesting, making you want to come back.

If people initially joined Twitter because that’s where the up-to-date conversations were happening, then Bluesky may benefit from becoming the new “it” place, supplanting X/Twitter’s former status as the place to be.

However, Bluesky could still go the way of other X competitors if it doesn’t maintain momentum and reach that critical mass. We’ve seen plenty of hype cycles before. Läs mer…

Two old Russian tankers, carrying thousands of tonnes of oil, have been severely damaged in a storm in the Kerch Strait.

The Volgoneft-212 tanker split in half after being hit by a large wave. Video footage shows the bow visible in the water and black oil surrounding the destruction. The second tanker, Volgoneft-239, then drifted into the same area shortly afterwards due to storm damage.

The Kerch Strait has served as a crucial logistical and transport hub for Russia since the second world war. In 2023, Ukraine attacked the Kerch Strait bridge, forcing Russia to rely on land routes to supply frontlines.

On December 16, Ukraine accused Moscow of breaking navigation safety rules by keeping the vessels at sea in such extreme weather and using old tankers. Reported by Tass, a Russian state-owned news agency, the two boats carried up to 9,200 tonnes of oil. According to satellite data, around 3,000 tonnes of heavy fuel oil has leaked into the Black Sea.

Marine oil spills are one of the most concerning sources of water pollution, destroying entire ocean habitats and the animals, plants and microbes living there.

When oil is mixed with water, the heavy components sink towards the seabed, leaving a thick residue on the water’s surface. The waves further mix the oil and water, which is increased in stormy weather. The oil can be contained in multiple ways – by using temporary floating barriers, burning the oil, using chemicals to disperse the oil or using boats to remove the oil from the surface.

Without containing the oil, it spreads further. By the morning of December 17, oil had already washed up on southern Russia’s Black Sea coast.

The thick oil slick on the water’s surface endangers the growth and development of marine life, both at sea and along the coastline. For example, oil sticks to a bird’s feathers, causing them to mat and separate, impairing the waterproofing of the feathers.

When birds preen their oil-covered feathers, they ingest the oil. Residents along the shores of the Kerch Strait in the Russian village of Volna have reported that oil-covered birds can no longer fly and have been forced to seek safety on the beaches.

Hidden costs

A similar incident occurred in this region in 2007 when about 1,300 tonnes of oil was spilt. Russia and Ukraine did not ask for international assistance to tackle the spills. However, in 2007 many international organisations helped with wildlife rescue efforts and on-coast cleaning operations, with more than 500 volunteers getting involved.

This week, more than 4,000 volunteers are cleaning a 30-mile stretch of shoreline near Anapa as the oil spill has not been contained.

Coordinated international efforts and vast resources are needed to restrict the oil spill and limit environmental damage. However, because of the conflict between Russia and Ukraine, it is still unknown whether international efforts will be conducted in the Russian-occupied area. And Ukraine cannot perform another clean-up operation like the one in 2007. The total costs from clean up operations, wildlife loss, loss of trade and tourism of the 2007 oil spill was estimated by the UN Environment Programme to be at least US$25 million (£20 million).

The environmental damage and loss of biodiversity are the hidden costs of conflict. Despite international clean-up efforts, 30,000 birds and countless fish died in the 2007 oil spill. There is no international effort to clean up the current spill, so the environmental effects could be much worse in the coming days.

Since 1992, the Black Sea has been monitored for pollution, restoration, protecting marine resources and subject to international environmental law under the Bucharest convention. However, environmental catastrophes continue in this area, not stopping with the conflict. This area affects many bordering countries, but the lasting ecological damage affects us all.

Regulators, conservation charities and environmental activists should monitor oil and gas activities and find areas where such business operations should be banned. Oil and gas firms should have updated emergency plans, and legal entities should run regular audits of these operations. The UN’s latest climate summit called for the urgent phasing out of fossil fuels. This environmental disaster need to accelerate the green energy transition and no longer take the environment for granted.

Don’t have time to read about climate change as much as you’d like?
Get a weekly roundup in your inbox instead. Every Wednesday, The Conversation’s environment editor writes Imagine, a short email that goes a little deeper into just one climate issue. Join the 40,000+ readers who’ve subscribed so far. Läs mer…

“Talk to me before you believe anything!”, Thomas Cromwell (played by Mark Rylance) tells his monarch and master Henry VIII (played by Damien Lewis) in a pivotal scene in the penultimate episode of the BBC adaptation of The Mirror and the Light, the 900-page third act in Hilary Mantel’s acclaimed Tudor drama. It is the last of their candid exchanges before the criticism from his enemies becomes deafening and Cromwell is seized, stripped of his offices of state and taken into custody.

For nearly a decade, Cromwell, has strived to speak the truth to his all-mighty monarch. It has set him apart from all those packed into the palace antechambers and pitching for a seat on the king’s council: aristocrats who were there by hereditary right, churchmen because of their office and favourites whose rise invariably was as fast as their fall.

Shrewd realpolitik was why Cromwell’s assent to the top seemed unstoppable and the reason Henry grew to like him. “By St Loy, this man has stomach, this man has gall!,” the now-ailing king recalls, in Peter Straughan’s script, of his first impression. Establishing the facts of the matter, however uncomfortable for whichever party, remained this self-taught lawyer’s stock-in-trade. But by 1540 Henry lacked the bodily strength and mental self-assurance to accept them anymore. “I have changed, Thomas. You, not so much.”

Looking for something good? Cut through the noise with a carefully curated selection of the latest releases, live events and exhibitions, straight to your inbox every fortnight, on Fridays. Sign up here.

“To retrieve history,” Mantel argued in her 2017 Reith lecture, “we need rigour, unsparing devotion and an impulse to scepticism.” Stomach and gall. Enough, Mantel urged, to unsettle the stereotypes, even to “de-centre” the “grand narrative”.

The determination Mantel found in Cromwell’s letters, to “decipher the bottom of their heart … if by any wisdom it may be drawn out” to politician Sir Thomas Wyatt in 1537 suggested that he could be a fresh guide to this period of Tudor history, long distorted by caricature and cliché. As sharp as the dagger secreted in his doublet and an outsider on the inside, through Cromwell’s eyes it might be refocused.

Mantel believed the novelist could best evoke an alternative point-of-view. “The records do … throw up some facts but they are not the whole truth.” Lived experience, she argued, lies in the “gaps, the erasures and silences” of the documented past, speeches unrecorded, thoughts unspoken. Fiction breathes them back into life, lifting the veil from the “vital”, “interior” view of lives long past.

Mantel’s conviction was so strong that she was inclined to criticise readers for clinging to “the first history they learn” and for their “unreasonable” refusal to commit to the novelist’s telling of it until they can be sure of its reliability. “I report the outer world faithfully,” she explained “but my chief concern is the interior drama of my characters’ lives.”

In their adaptation Straughan and director Peter Kosminsky have followed Mantel’s manifesto to the letter. They present the “outer world” with precision.

Unlike any of his predecessors, Kosminky has taken immense pains to locate the action in landscapes and environments which the historical figures would have recognised. His choices are clever, including Gloucester Cathedral, surely the most complete and unaltered Benedictine cloister in Britain, standing in for the lost abbey of Shaftesbury and Horton Court in Gloucestershire for Cromwell’s city of London chambers at the Austin Friary, which, in spite of its religious status, is known to have looked like a cluster of town houses.

Casting director Robert Sterne has peopled the scene with performers uncannily close in age and aspect to the figures they play. Thomas Brodie Sangster (Rafe Sadler) and Harry Melling (Thomas Wriothesley), both in their 30s, capture their subjects perfectly. Damien Lewis still seems to walk in Henry’s now halting footsteps, although the hiatus since the 2015 series means he is three years the king’s senior. Only Timothy Spall (Thomas Howard) is jarring. He is 67, as Howard was in 1540, but squat, jowly and with a one-note anger, which makes him less like a noble duke and more like Alice through the Looking Glass’s Queen.

Damien Lewis is three years older than Henry VIII was during the period the series covers but does a good job.
BBC/Playground Entertainment/Nick Briggs

But in spite, or perhaps because, of the studied skill of this practised team, this visualisation does expose the tensions – in fact, downright contradictions – in Mantel’s treatment of the past. In her Reith Lectures she declared: “Don’t lie, don’t go against known facts. Historical truth cuts against the storyteller’s instinct. Your characters are never how or where you’d like them to be.” Yet Straughan’s adept précis of the 900-page book shows how often she shrugged off her own counsel.

The visible, vocal presence of Cardinal Thomas Wolsey (Jonathan Pryce), dead a full six years before the events of episode one unfolded, is conspicuously clumsy, muddling any viewer confronting the subject for the first time. Of course, it is a device, but if the “outer world” seems unstable how can they make sense of the “interior drama”? In Cromwell’s marriage proposal to the cardinal’s cast-off daughter, Dorothy Clancy (episode two) Mantel places her characters where she would like them to be and to say what she would like them to say.

Her “going against known facts” is less troubling than her narrow line-of-sight, which the clipped script and slick camerawork set in sharp relief. The historical dramas of Cromwell’s last years in power were armed rebellion (the Pilgrimage of Grace) and the greatest displacement of people and livelihoods since the Norman Conquest (the dissolution of the monasteries), a stop-start process which de-stabilised the Tudor regime as much as its subjects. In Mantel’s story, they are little more than noises out of frame.

The climate in court and country became so febrile in the face of these episodes because positions on them in every part of society were uncertain even, perhaps especially, in the mind of the king himself. Thomas Cromwell knew this better than any other contemporary witness. Mantel, whatever she claimed to the contrary, maintained the “grand narrative” of Catholic versus Protestant, traditional aristocrat versus modernising commoner. Like King Henry, she did not keep listening to her subject for long enough. Läs mer…