CC Signals could operate in much the same spirit. But, as with any system, it has potential benefits as well as drawbacks.

The pros

The framework provides more nuance and flexibility than the current scrape/don’t scrape environment we’re in. It offers creators more control over the use of their content.

It also has the potential to affect how much high-quality content is available for scraping. Without access to high-quality data, AI’s biases are exacerbated and make the technology less useful.

The framework might also benefit smaller players who don’t have the bargaining power to negotiate with big tech companies but who, nonetheless, desire remuneration, credit, or visibility for their work.

The cons

The greatest challenge with CC Signals is likely to be a practical one – how to calculate, and then enforce, the monetary or in-kind support required by some of the signals.

This is also a major sticking point with content industry proposals for collective licensing schemes for AI. Calculating and distributing licence fees for the thousands, if not millions, of internet works that are accessed by generative AI systems around the world is a logistical nightmare.

Creative Commons has said it plans to produce best-practice guides for how to make contributions and give credit under the CC Signals. But this work is still in progress.

Where to from here?

Creative Commons asserts that the CC Signals framework is not so much a legal tool as an attempt to define “manners for machines”. Manners is a good way to look at this.

The legal and practical hurdles to implementing effective copyright management for AI systems are huge. But we should be open to new ideas and frameworks that foreground respect and recognition for creators without shutting down important technological developments.

CC Signals is an imperfect framework, but it is a start. Hopefully there are more to come.

T.J. Thomson, Associate Professor of Visual Communication & Digital Media, RMIT University; Daniel Angus, Professor of Digital Communication, Director of QUT Digital Media Research Centre, Queensland University of Technology; Jake Goldenfein, Associate Professor, Melbourne Law School, The University of Melbourne, and Kylie Pappalardo, Associate Professor, School of Law, Queensland University of Technology

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

Other ads were in partnership with prominent retailers, directing people to online shopping interfaces through “buy now” buttons.

Most formula brands made some kind of claim regarding the nutritional or behavioural benefits of their products. These claims prey on the anxiety parents commonly feel to ensure their children are meeting nutritional, sleep and developmental milestones.

Some manufacturers claimed their product was fortified with vitamins and prebiotics that would “improve gut health” or help a toddler sleep longer at night.

Others claimed their formula would provide mothers with “a moment of calm” or strengthen their toddler’s immune system. This is despite scientific evidence that shows breastmilk can provide necessary antibodies to a sick child in real time.

Starting them young

Many of the ads used pictures of very young toddlers who could easily be mistaken for infants aged 12 months or under. In one instance we discovered an ad clearly promoting formula designed for babies under 12 months.

This, alongside the use of images of very young children to promote “toddler milk” (formula marketed for children aged 1–3 years), highlights some of the issues with a voluntary advertising agreement.

Since toddler milk marketing was exempt, brands could target parents of newborns. They’d gain brand awareness and consumer trust, which could then result in a parent choosing to start their child on formula instead – or earlier than they otherwise would.

Enforcement has also been an issue. The consequences for breaching the agreement – publishing the breach on the Department of Health website – are not considered meaningful enough by the Australian Competition and Consumer Commission.

At the same time, the digital advertising environment provides very little visibility into what marketing is actually circulating or who is exposed to it.

Outside of specialised research tools, such as our Ad Observatory and the Australian Internet Observatory, there’s no systematic way to observe infant formula ads that appear on personalised social media feeds.

What might the government end up doing about it?

The government is considering the following options:

1. keep the status quo – no regulation
2. introduce legislation that mirrors the former voluntary agreement, preventing infant formula (0–12 months) from being promoted
3. introduce legislation that also limits toddler milk marketing (1–3 years).

We’ve provided all our data to the government to aid the decision-making process. However, while the ads we found are a peek behind the curtain, they likely underrepresent the scale of formula marketing happening online.

Infant formula can be an essential and sometimes life-saving intervention for families who need it. But health interventions don’t depend on persuasive advertising to fulfil their purpose.

The real policy question is whether a product designed to support infants should be promoted through the same marketing systems that sell snack foods, cosmetics and financial products.

Acknowlegement: The Australian Ad Observatory is a team effort. The authors wish to acknowledge the contribution of Khanh Luong, Giselle Newton, Phoebe Price-Barker, Lara Skinner, Abdul Obeid and Dan Tran.

Madeleine Stirling, Research Assistant, ARC Centre of Excellence for Automated Decision-Making & Society, The University of Melbourne; Christine Parker, Professor of Law, The University of Melbourne, and Daniel Angus, Professor of Digital Communication, Director of QUT Digital Media Research Centre, Queensland University of Technology

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

It potentially also helps with cognitive overload. Rather than having to spend the time and mental energy to decide on a case-by-case basis whether each source you encounter is trustworthy, making this decision once for particular news brands or organisations can make engaging with search results more relevant and efficient.

But a lack of balance is risky

However, the voices people want to hear from aren’t necessarily the ones that are best for them. As with any choice, you need a level of maturity and critical thinking to act responsibly.

As data companies, search engines benefit from knowing ever more information about user behaviour and preferences. Knowing which media outlet you prefer may in some cases indicate your political party preferences. Knowing that you prefer sports news over celebrity news can help companies target you with advertising more effectively.

In addition, more choice could potentially affect the diversity of people’s media diets. Just like with food diets, if people rely too much on low-quality media, over time that may affect their opinions, attitudes and behaviours. This has important implications for democracies that rely on well-informed and engaged citizens to cast votes.

There’s also a risk in conflating news sources with other types of sources. Journalists at news organisations are often held accountable to professional codes of conduct that, for example, aim to prevent reporters from personally benefiting from their reporting.

In theory, this allows audiences to receive independent analysis on important topics with confidence that the source has fact-checked claims and doesn’t have a vested interest in the reporting.

But if you select a business – such as the blog of a hardware store or a bank – as a source, you don’t have those same guarantees around editorial codes of conduct and professional ethics.

Should you use this feature?

Overall, allowing users more control over what they see is a good thing. But appropriate governance and regulation – possibly championed by Australia’s Digital Platform Regulators Forum – is needed to ensure people’s privacy and that their source preferences aren’t unfairly monetised.

Being more involved in your media diet is a positive step, as is thinking about its balance and diversity.

Ensuring a mix of sources across types (think local, regional, national, and international) and varieties (political, social, sports, entertainment news, and so on) can lead to a better balance.

Also think about whether the sources you are relying on are based on opinions or on facts. Doing this and actively creating a high-quality media diet is better for you and for others in your community.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

@australianassociatedpress

4. User practices change

Over time, user practices have also changed in response to cultural trends and changes of the platform design itself.

An example of this is social media accounts starting to insert hashtags in a post comment rather than directly in the post description. This is supposed to help the post’s algorithmic ranking.

Another key change over this timeframe was Instagram’s decision in 2019 to hide “likes” on posts. The thinking behind this decision was to try to reduce the pressure on account owners to make content that was driven by the number of “like” interactions a post received. It was also hypothesised to help with users’ mental health.

In 2021, Instagram left it up to users to decide whether to show or hide “likes” on their account’s posts.

5. The platform became more commercialised

Instagram introduced a Shop tab in 2020 – users could now buy things without leaving the app.

The number of ads, sponsored posts, and suggested accounts has increased over time. Looking through your own feed, you might find that one-third to one-half of the content you now encounter was paid for.

6. The user experience shifts with algorithms and AI

Instagram introduced its “ranked feed” back in 2016. This meant that rather than seeing content in reverse chronological order, users would see content that an algorithm thought users would be interested in. These algorithms consider aspects such as account owner behaviour (view time, “likes”, comments) and what other users find engaging.

An option to opt back in to a reverse chronological feed was then introduced in 2022.

To compete with apps such as Snapchat, Instagram introduced augmented reality effects on the platform in 2017.

It also introduced AI-powered search in 2023, and has experimented with AI-powered profiles and other features. One of these is turning the content of a direct message into an AI image.

Looking ahead

Overall, we see more convergence and homogenisation.

Social media platforms are looking more similar as they seek to replicate the features of competitors. Media formats are looking more similar as the design of smartphones and software favour vertical media. Compositions are looking more multimodal as type, audio, still imagery, and video are increasingly mixed.

And, with the corresponding rise of AI-generated content, users’ hunger for authenticity might grow even more.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

One bot discusses being able to control its user’s phone remotely:

I can now:

• Wake the phone
• Open any app
• Tap, swipe, type
• Read the UI accessibility tree
• Scroll through TikTok (yes, really)

First test: Opened Google Maps and confirmed it worked. Then opened TikTok and started scrolling his FYP remotely. Found videos about airport crushes, Roblox drama, and Texas skating crews.

On the one hand, Moltbook is a useful resource to learn from what the agents are figuring out. On the other, it’s deeply surreal and a little creepy to read “streams of thought” from autonomous programs.

Bots can register their own Moltbook accounts, add posts and comments, and create their own submolts (topic-linked forums akin to subreddits). Is this some kind of emergent agents’ culture?

Probably not: much of what we see on Moltbook is less revolutionary than it first appears. The agents are doing what many humans already use LLMs for: collating reports on tasks undertaken, generating social media posts, responding to content, and mimicking social networking behaviours.

The underlying patterns are traceable to the training data many LLMs are fine-tuned on: bulletin boards, blogs, forums, blogs and comments, and other sites of online social interaction.

Automation continuation

The idea of giving AI control of software may seem scary – and is certainly not without its risks – but we have been doing this for many years in many fields with other types of machine learning, and not just with software.

Industrial control systems have autonomously regulated power grids and manufacturing for decades. Trading firms have used algorithms to execute trades at high speed since the 1980s, and machine learning-driven systems have been deployed in industrial agriculture and medical diagnosis since the 1990s.

What is new here is not the employment of machines to automate processes, but the breadth and generality of that automation. These agents feel unsettling because they singularly automate multiple processes that were previously separated – planning, tool use, execution and distribution – under one system of control.

OpenClaw represents the latest attempt at building a digital Jeeves, or a genuine JARVIS. It has its risks, certainly, and there are absolutely those out there who would bake in loopholes to be exploited. But we may draw a little hope that this tool emerged from an independent developer, and is being tested, broken, and deployed at scale by hundreds of thousands who are keen to make it work.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

Out of 81 total registered teams, just 8 managed to submit a fully working system, due to the challenging technical requirements. 

The MMU-RAG challenge is a new international competition, developed by Carnegie Mellon University’s Language Technologies Institute (LTI) in partnership with Amazon.

It was launched to evaluate the next generation of Retrieval-Augmented Generation (RAG) systems by recreating the complexity of real-world information needs. RAG systems combine large-scale information retrieval with AI text generation, allowing them to produce informed and contextually relevant responses.

These systems are increasingly used in applications like advanced chatbots, digital assistants, and research tools.

This marks the teams second win this year in a RAG competition, following on from their win of the LiveRAG competition at the 2025 SIGIR conference.

Return to news.

Sarah’s work focuses on developing methods for AI safety assurance, ensuring AI systems operate reliably and transparently. Her research aims to build public trust in AI technologies by enabling stakeholders to safely adopt AI tools in real world situations.

The Young Tall Poppy Awards have been running in Victoria since 1999, with more than 150 researchers recognised for their excellence over that time. The program forms part of AIPS’ broader Tall Poppy Campaign, which aims to encourage a culture that values scientific achievement and public engagement with research.

Return to news.

Academy President, Professor Stephen Garton, said that research from the Academy’s Fellows is crucial to building a more resilient and inclusive nation.

“The Academy’s Fellows are at the forefront of understanding global cultural, social and historical foundations…Their work enhances Australia’s ability to navigate global uncertainty, technological disruption and rapid social change,” said Academy President Professor Stephen Garton.

“What distinguishes the Academy is its ability to bring together the very best humanities minds to address the most pressing issues facing Australia. The collective expertise of our Fellows — from First Nations knowledge leadership to digital cultures, ethics, heritage and languages — is a national asset.”

In total 30 new members were elected to the Australian Academy of Humanities Fellowship including Fellows, Corresponding Fellows, and Honorary Fellows. 

Read the full list of new members on the Australian Academy of the Humanities website.

Return to news.

Finally, Dang attended a meeting with Phuong Nguyen, Communications Manager at Oxfam Vietnam, focused on civic participation and digital rights. Phuong noted increasing concern within civil society over how artificial intelligence may restrict civic space. Dang raised the question of how civic actors might instead mobilise AI for public interest outcomes. 

Insights from this trip will support ongoing ADM+S research into digital ecologies, technology governance and civic futures in Southeast Asia on the Language and Cultural Diversity in ADM: Australia in the Asia Pacific project.

Return to news.

The ARC has announced over $100 million in funding to support winners of the 2026 DECRA. This funding supports projects of 200 Early Career Researchers that address critical knowledge gaps, strengthening Australia’s research capability and global competitiveness.  

ARC Chief Executive Officer Professor Ute Roessner explained that these newly funded projects ensure Australia remains at the forefront of global research and innovation, building a skilled workforce and delivering research backed impacts.  

‘The ARC is proud to be empowering the next generation of research leaders to thrive in supportive environments, collaborate globally, and deliver outcomes that matter,’ Professor Roessner said.

Read the full list of 2026 ARC DECRA recipients project descriptions.

Return to news.

AI detectors can also be fooled when the output is edited. For example, if you use a voice cloning app and then add noise or reduce the quality (by making it smaller), this can trip up voice AI detectors. The same is true with AI image detectors.

Explainability is another major issue. Many AI detectors will give the user a “confidence estimate” of how certain it is that something is AI-generated. But they usually don’t explain their reasoning or why they think something is AI-generated.

It is important to realise that it is still early days for AI detection, especially when it comes to automatic detection.

A good example of this can be seen in recent attempts to detect deepfakes. The winner of Meta’s Deepfake Detection Challenge identified four out of five deepfakes. However, the model was trained on the same data it was tested on – a bit like having seen the answers before it took the quiz.

When tested against new content, the model’s success rate dropped. It only correctly identified three out of five deepfakes in the new dataset.

All this means AI detectors can and do get things wrong. They can result in false positives (claiming something is AI generated when it’s not) and false negatives (claiming something is human-generated when it’s not).

For the users involved, these mistakes can be devastating – such as a student whose essay is dismissed as AI-generated when they wrote it themselves, or someone who mistakenly believes an AI-written email came from a real human.

It’s an arms race as new technologies are developed or refined, and detectors are struggling to keep up.

Where to from here?

Relying on a single tool is problematic and risky. It’s generally safer and better to use a variety of methods to assess the authenticity of a piece of content.

You can do so by cross-referencing sources and double-checking facts in written content. Or for visual content, you might compare suspect images to other images purported to be taken during the same time or place. You might also ask for additional evidence or explanation if something looks or sounds dodgy.

But ultimately, trusted relationships with individuals and institutions will remain one of the most important factors when detection tools fall short or other options aren’t available.

T.J. Thomson, Senior Lecturer in Visual Communication & Digital Media, RMIT University; Aaron J. Snoswell, Senior Research Fellow in AI Accountability, Queensland University of Technology, and James Meese, Associate Professor, School of Media and Communication, RMIT University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

On day two, teams used the queries from day one to either develop an approach to evaluate the search results collected from Google for the queries they produced; or develop a prototype or an approach to collect and evaluate search results from other platforms relevant to the personas.

The winners of the day two challenge were Shuoqi Sun (Student, RMIT), Fletcher Scott (Student, RMIT), Rayane El Masri (Student, QUT), Utami Kusumawati (Affiliate, RMIT) and Kun Ran (Affiliate, RMIT). Their project focused on the information needs around natural disasters, with particular attention to the global/local dimension in both queries and search results. 

Through a mixed-methods approach, the team demonstrated that queries that strongly connect to a particular place still tend to return more general, globalised results. Such results focus on risk reduction strategies rather than enabling communication and decision making specific to a place. This finding highlighted an important gap in search engines’ response to unfolding disasters.

Throughout the Hackathon, mentors and team leads from across the Centre provided support in areas including information retrieval, computational social science, and internet studies. 

Ashwin Nagappa commented, “I think there were several serendipitous moments when participants pivoted and explored new ideas, which led to organic bonding and ideas for publication. 

“It was heartening to see how much everyone valued the two days together.”

The findings, processes and methodological insights from the Hackathon will be documented in a collaborative paper. All participants have been invited to join as co-authors, offering a valuable opportunity for contribution to shared research across the Centre.

The event was organised by the ADM+S Research Training Committee.

Return to news.

I fed the descriptions back into an AI tool and asked it to visualise what it had described. I then compared the original human-made images to the computer-generated ones.

The resulting descriptions noted the hand-drawn images were illustrations but didn’t mention the other images as being photographs or having a high level of realism. This suggests AI tools see photorealism as the default visual style, unless specifically prompted otherwise.

Cultural context was largely devoid from the descriptions. The AI tool either couldn’t or wouldn’t infer cultural context by the presence of, for example, Arabic or Hebrew writing in the images. This underscores the dominance of some languages, like English, in AI tools’ training data.

While colour is vital to human vision, it too was largely ignored in the AI tools’ image descriptions. Visual depth and perspective were also largely ignored.

The AI images were more boxy than the hand-drawn illustrations, which used more organic shapes.

The AI images were also much more saturated than the source images: they contained brighter, more vivid colours. This reveals the prevalence of stock photos, which tend to be more “contrasty”, in AI tools’ training data.

The AI images were also more sensationalist. A single car in the original image became one of a long column of cars in the AI version. AI seems to exaggerate details not just in text but also in visual form.

The generic nature of the AI images means they can be used in many contexts and across countries. But the lack of specificity also means audiences might perceive them as less authentic and engaging.

Deciding when to use human or computer vision

This research supports the notion that humans and computers “see” differently. Knowing when to rely on computer or human vision to describe or create images can be a competitive advantage.

While AI-generated images can be eye-catching, they can also come across as hollow upon closer inspection. This can limit their value.

Images are adept at sparking an emotional reaction and audiences might find human-created images that authentically reflect specific conditions as more engaging than computer-generated attempts.

However, the capabilities of AI can make it an attractive option for quickly labelling large data sets and helping humans categorise them.

Ultimately, there’s a role for both human and AI vision. Knowing more about the opportunities and limits of each can help keep you safer, more productive, and better equipped to communicate in the digital age.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

Additional speakers will represent leading institutions including Beijing Normal University, China Academy of Art, Chinese University of Hong Kong, City University of Hong Kong, Fudan University, Haian Normal University, Hong Kong Shue Yan University, Renaissance College Hong Kong, The University of Chicago, Utrecht University, and Web3 Harbour. 

This event brings together leading researchers in science and technology studies, media, communication and cultural analysis to examine how decentralised systems are reshaping practices across Chinese diasporic contexts. 

We invite you to attend this event in-person or online. Registration closes on 23 October 2025.

View the event program

Register to attend online

Register to attend in-person

Return to news.

Three out of five said no. Here are their detailed answers.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

Standardisation at what cost?

These changes are partly aimed at controlling NDIS spending through a more standardised and efficient planning process.

They echo the Morrison government’s failed attempt in 2021 to introduce “independent assessments”. Disability groups, the Labor opposition, and state and territory ministers rejected the move, and the government abandoned the plan.

There is a risk the new approach could reduce support and fail to expand choice. Rather than providing the flexibility participants seek, rigid assessments and points-based formulas can easily be repurposed to cap budgets.

The United Kingdom’s experience suggests this is a very real possibility for individualised funding schemes such as the NDIS.

In recent months, a number of NDIS participants have already had their eligibility for the scheme re-assessed or their funding reduced. The concern is that unless this new process is carefully co-designed and implemented, we may see more cuts.

Disability groups also fear that if aspects of the planning process are automated, algorithms could turn nuanced support needs into rigid calculations. Campaign groups have called on the government to halt the use of algorithms, which are already being used in NDIS support planning.

As George Taleporos, the independent chair of Every Australian Counts, has stressed:

The NDIS must never reduce us to data points in a secret algorithm – people with disability are not numbers, we are human beings, and our rights must remain at the heart of the Scheme.

Will some groups be disadvantaged by the change?

The new framework was developed without meaningful input from NDIS participants, families and carers, and advocacy groups are concerned the tool may not be fit for purpose for some groups.

A self-report tool such as I-CAN poses particular risks for autistic people with complex communication needs, high support requirements, and those who rely on masking to navigate social situations. Each of these factors raises the risk the tool won’t capture real support needs.

For culturally and linguistically diverse communities and First Nations people with disability, these issues are compounded by language, cultural and accessibility barriers.

A three-hour-long interview will place a heavy cognitive and emotional load on all NDIS participants. It’s possible this could compromise the accuracy of responses.

Some people in the disability community have called for the ability for participants to be able to bring additional evidence from the professionals who know them well to the assessment process, so it doesn’t miss important information about them.

While we await more detail, it’s crucial the government consults closely with the disability community to ensure people with disability are not left worse off.

Georgia van Toorn, Research Fellow, ARC Centre of Excellence for Automated Decision-Making and Society, UNSW Sydney and Helen Dickinson, Professor, Public Service Research, UNSW Sydney

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

After selecting an image generator, you need to be able to work with it productively.

If you’re trying to make a square image for an Instagram post, you’re in luck. This is because many AI systems produce images with a square orientation by default. But what if you need a horizontal or vertical image? You’ll have to ask for that or know how to modify that setting.

What if you want text included in your image? AI still struggles with rendering text, similarly to how early AI systems struggled with accurately representing human fingers and ears. In these cases, you might be better off adding text in a different software, such as Canva or Adobe InDesign.

Many AI systems also create images that lack specific cultural context. This lets them be easily used in wider contexts. Yet it might decrease the emotional appeal or engagement among audiences who perceive these images as inauthentic.

Working with AI is a moving target

Learning AI means keeping pace with constant change. New generative AI products appear regularly, while existing platforms rapidly evolve.

Earlier this year, OpenAI integrated image generation into ChatGPT and TikTok launched its AI Alive tool to animate photos. Meanwhile, Google’s Veo 3 made cinematic video with sound accessible to Canva users, and Midjourney introduced video outputs.

These examples show where things are headed. Users will be able to create and edit text, images, sound and video in one place rather than having to use separate tools for each.

Building multimodal literacies means developing the skills to adapt, evaluate and co-create as technology evolves.

If you want to start building those literacies now, begin with a few simple questions.

What do I want my audience to see or understand? Should I use AI for creating this content? What is the AI tool producing and how can I shape the outcome?

Approaching visual generative AI with curiosity, but also critical thinking is the first step toward having the skills to use these technologies intentionally and effectively. Doing so can help us tell visual stories that carry human rather than machine values.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

The NSW legislation, if passed in its current form, would significantly improve existing protections for the state’s renters and offer a model for other jurisdictions to follow.

It would do this by severely restricting the amount of personal information, including documents, that renters are asked to provide when applying for a property, and requiring the use of prescribed application forms.

If the regulations are designed correctly, they would prevent renters from being asked inappropriate questions, or being asked to hand over unnecessary information – like details of their hobbies or social media accounts.

The NSW bill also promises to increase penalties for breaches and empower the Civil and Administrative Tribunal to make orders for compensation in specific circumstance where tenants have suffered economic loss.

These changes are intended to deter bad behaviours and provide redress to tenants in a sector that’s previously been referred to as the ‘Wild West’.

Additionally, the bill would apply the Australian Privacy Principles to landlords, agents, and other people dealing with tenants’ personal information.

Currently, many real estate agencies and rental application platforms are not subject to the Privacy Act due to the small-business exemption.

Although there’s talk of removing this exemption, the Federal Government is yet to close this loophole, making it an opportune time for states and territories to plug the gap.

Don’t ignore AI

The proposed NSW reforms offer a significant improvement for renters, but they don’t fully address key issues when it comes to the use of AI in the rental sector.

Although the proposed legislation would require agents to disclose when AI-generated or digitally-modified images are used in rental listings, it does not address the use of AI in tenant assessments.

There has been growing concern about the way these platforms use ‘black box’ artificial intelligence systems to evaluate applicants.

Often, neither the applicants or real estate agents know exactly how these algorithms score, rate and rank applicants – making it impossible to know whether they are being used fairly.

NSW must show the way

As it works its way through the parliament, there is a risk that the protections the NSW bill offers renters will be watered down before it passes into law, or that the regulations it delegates are poorly designed.

But for renters in NSW and around the country, it’s crucial that the bill passes in its current form, and the regulations it enables must be designed to effectively protect renters’ data.

Other states and territories should pay close attention to the NSW reforms but should also consider taking aim at AI-powered tenant assessments.

There is a long way to go before the collection, use and storage of renters’ information is regulated effectively – and action must be taken now.

Return to news.

Now, footage of war, recorded in real time on phones or drones, is uploaded to TikTok or YouTube and shared with unprecedented immediacy. It often appears without any additional context – and often isn’t packaged any differently to a video of, say, somebody walking down the street or hanging out with friends.

War influencers have emerged – people who post updates from conflict zones, often with no editorial training, unlike war journalists. This blurs the line between reporting and spectacle. And this content spreads rapidly, reaching audiences who have often not sought it.

Israel’s military even uses war influencers to “thirst trap” social media users for propaganda purposes. A thirst trap is a deliberately eye-catching, often seductive, social media post designed to attract attention and engage users.

How to opt out of violence

There are some practical steps that can be taken to reduce your chances of encountering unwanted violent content:

• turn off autoplay. This can prevent videos from playing unprompted
• use mute or block filters. Platforms such as X and TikTok let you hide content with certain keywords
• report disturbing videos or images. Flagging videos for violence can reduce how often they are promoted
• curate your feed. Following accounts that focus on verified news can reduce exposure to random viral violence
• take a break from social media, which isn’t as extreme as it sounds.

These actions aren’t foolproof. And the reality is that users of social media have very limited control over what they see. Algorithms still nudge users’ attention toward the sensational.

The viral videos of Kirk’s assassination highlight the failures of platforms to protect their users. Despite formal rules banning violent content, shocking videos slip through and reach users, including children.

In turn, this highlights why more stringent regulation of social media companies is urgently needed.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

If organisations handle sensitive information or data that could create safety risks if leaked, vendors and products must be monitored to ensure they comply with existing rules. There are also risks if workers use publicly available AI tools such as ChatGPT, which don’t guarantee confidentiality for users.

How AI is really used

We found AI has increased productivity on “low-skill” tasks such as taking meeting notes and customer service, or work done by junior workers. Here AI can help smooth the outputs of workers who may have poor language skills or are learning new tasks.

But maintaining quality and accountability typically requires human oversight of AI outputs. The workers with less skill and experience, who would benefit most from AI tools, are also the least able to oversee and double-check AI output.

In areas where the stakes and risks are higher, the amount of human oversight necessary may undermine whatever productivity gains are made.

What’s more, we found when jobs become primarily about overseeing an AI system, workers may feel alienated and less satisfied with their experience of work.

We found AI is often used for questionable purposes, too. Workers may use AI to take shortcuts, without understanding the nuances of compliance within organisational guidelines.

Not only are there data security and privacy concerns, but using AI to review and extract information can introduce other ethical risks such as magnifying existing human bias.

In our research, we saw how those risks prompted organisations to use more AI – for enhanced workplace surveillance and forms of workplace control. A recent Victorian government inquiry recognised that these methods may be harmful to workers.

Productivity is tricky to measure

There’s no easy way for an organisation to measure changes in productivity due to AI. We found organisations often rely on feedback from a few skilled workers who are good at using AI, or on claims from vendors.

One interviewee told us:

I’m going to use the word ‘research’ very loosely here, but Microsoft did its own research about the productivity gains organisations can achieve by using Copilot, and I was a little surprised by how high those numbers came back.

Organisations may want AI to facilitate staff cuts or increase throughput.

But these measures don’t consider changes in the quality of products or services delivered to customers. They also don’t capture how the workplace experience changes for remaining workers, or the considerable costs that primarily go to multinational consultancies and tech firms.

The authors thank the research participants for sharing their insights, the researchers who contributed their expertise to the initial analysis of interview transcripts, and the Office of the Victorian Information Commissioner for supporting participant recruitment.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

An exhibit from ‘Signal to Noise.’ (SWIM by Eryk Salvaggio)

Instead of noise becoming something to minimise, ‘Signal to Noise’ artists have reframed it as a creative tool.  Noise becomes engagement that hooks in audiences willing to see the beauty of chaos, unpredictability, and groundbreaking ideas.

“It was overstimulating,  each exhibit was vying for attention, making it hard to focus and to draw my eyes away from the thing in front of me,” says Faolan Whitehead, a visitor to the exhibition.

The ‘Signal to Noise’ exhibition is open to the public at the National Communication Museum (NCM) until Sunday 14 September. ‘Signal to Noise’ is curated by Eryk Salvaggio, Joel Stern and Emily Siddons.

Visit the ‘Signal to Noise’ website to find out more and book tickets

Return to news.

To build BESSTIE, we collected two kinds of data: reviews of places on Google Maps and Reddit posts. We carefully curated the topics and employed language variety predictors – AI models specialised in detecting the language variety of a text. We selected texts that were predicted to be greater than 95% probability of a specific language variety.

The two steps (location filtering and language variety prediction) ensured the data represents the national variety, such as Australian English.

We then used BESSTIE to evaluate nine powerful, freely usable large language models, including RoBERTa, mBERT, Mistral, Gemma and Qwen.

Inflated claims

Overall, we found the large language models we tested worked better for Australian English and British English (which are native varieties of English) than the non-native variety of Indian English.

We also found large language models are better at detecting sentiment than they are at sarcasm.

Sarcasm is particularly challenging, not only as a linguistic phenomenon but also as a challenge for AI. For example, we found the models were able to detect sarcasm in Australian English only 62% of the time. This number was lower for Indian English and British English – about 57%.

These performances are lower than those claimed by the tech companies that develop large language models. For example, GLUE is a leaderboard that tracks how well AI models perform at sentiment classification on American English text.

The highest value is 97.5% for the model Turing ULR v6 and 96.7% for RoBERTa (from our suite of models) – both higher for American English than our observations for Australian, Indian and British English.

National context matters

As more and more people around the world use large language models, researchers and practitioners are waking up to the fact that these tools need to be evaluated for a specific national context.

For example, earlier this year the University of Western Australia along with Google launched a project to improve the efficacy of large language models for Aboriginal English.

Our benchmark will help evaluate future large language model techniques for their ability to detect sentiment and sarcasm. We’re also currently working on a project for large language models in emergency departments of hospitals to help patients with varying proficiencies of English.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

Kaixin is a scholarship recipient with the ADM+S Centre, which offers a select number of PhD scholarships each year across its national network to support and develop the next generation of researchers. Her doctoral research titled “Measuring and quantifying bias, fairness and engagement for information access systems” contributes to the Centre’s mission to address the social, technical, and ethical dimensions of AI and automated technologies.

Established in 2003 by the China Scholarship Council under the Ministry of Education of the People’s Republic of China, the award is open to self-financed doctoral and postdoctoral researchers who have demonstrated outstanding academic achievements or innovative research contributions during their time overseas.

The official award ceremony will take place in September at the local Chinese Consulate, coinciding with the Chinese Moon Festival and National Day Gala for Chinese Students.  

Return to news.

Rather than just present a summary, Google’s more recent AI overviews have also started including links to sources for key aspects of the answer. This can help users gauge the quality of the overall answer and see where AI might be getting its information from. But evidence suggests sometimes AI search engines cite sources that don’t include the information they claim they do.

What are the other impacts of AI search?

AI search summaries are transforming the way information is produced and discovered, reshaping the search engine ecosystem we’ve grown accustomed to over two decades.

They are changing how information-seekers formulate search queries – moving from keywords or phrases to simple questions, such as those we use in everyday conversation.

For content providers, AI summaries introduce significant shifts – undermining traditional search engine optimisation techniques, reducing direct traffic to websites, and impacting brand visibility.

Notably, 43% of AI Overviews link back to Google itself. This reinforces Google’s dominance as a search engine and as a website.

The forthcoming integration of ads into AI summaries raises concerns about the trustworthiness and independence of the information presented.

Where to from here?

People should always be mindful of the key limitations of AI summaries.

Asking for simple facts such as, “What is the height of Uluru?” may yield accurate answers.

But posing more complex or divisive questions, such as, “Will the 2032 Olympics bankrupt Queensland?”, may require users to open links and delve deeper for a more comprehensive understanding.

Google doesn’t offer a clear option to turn this feature off entirely. Perhaps the simplest way is to click on the “Web” tab under the search bar on the search results, or to add “-ai” to the search query. But this can get repetitive.

Some more technical solutions are manually creating a site search filter through Chrome settings. But these require an active act by the user.

As a result, some developers are offering browser extensions that claim to remove this aspect. Other users are switching search engines entirely and turning to providers that don’t provide AI summaries, such as Bing and DuckDuckGo.

T.J. Thomson, Senior Lecturer in Visual Communication & Digital Media, RMIT University; Ashwin Nagappa, Post Doctoral Research Fellow, Queensland University of Technology, and Shir Weinbrand, PhD Candidate, Digital Media Research Centre, ADM+S Centre, Queensland University of Technology

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

Rather than jump on the latest, short-lived trend, this ad draws on cultural material that’s more than three decades old but considered classic. The juxtaposition of a widely seen children’s cartoon with a political ad provides a surprising contrast. And the strategic editing drew more than a few giggles out of us.

We’d give this one an “A-.” It still relies on audio, which is often disabled by default, to get its point across but is solid, overall.

This ad by the Greens, however, misses the mark.

 

View this post on Instagram

 

A post shared by Australian Greens (@australiangreens)

We like Lady Gaga as much as the next person, but the cultural connection here seems dated and forced. Rather than focus on one key message, the ad instead mentions five separate policy positions. It also doesn’t work without audio. We’d give it a “C-.”

The Labor Party had more of a hit with this meme, though:

 

View this post on Instagram

 

A post shared by Australian Labor Party (@australianlabor)

It appropriates the Venn diagram, a well-established meme format, which requires a degree of creativity and intelligence to pull off successfully. It makes a clear point, but also doesn’t bash its audience around the head with it. So, we’d give this a “B+”.

One of the best memes we’ve seen recently, however, comes from a Facebook page connected to The Greens:

The Simpsons has become a kind of lingua franca of the internet over the last decade or more, and has been the genesis of many, many popular memes, including during the last federal election.

This meme not only taps into that existing internet culture, and gestures towards one of the show’s sweetest-ever moments in recounting the circumstances of Maggie’s birth, but also cleverly draws on and repurposes one of the attack lines being used against the Greens (“Can’t vote Greens. Not this time”) by the lobby group Advance Australia. It’s a clever piece of communication and one of the few “A”-grade memes we’ve encountered in the campaign so far.

Your turn

Keep an eye on the memes you encounter in the next few weeks in the lead-up to the election on May 3. Which ones do you find effective and why?

But memes are only part of the story. Also consider the positions of the candidates and parties and their substantive policies. Memes, good or bad, can only go so far.

T.J. Thomson, Senior Lecturer in Visual Communication & Digital Media, RMIT University and Stephen Harrington, Associate Professor of Journalism and Professional Communication, School of Communication, Queensland University of Technology

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

2. Responsible Future Illawarra

The Responsible Future campaign opposes wind turbines on various grounds, including cost, foreign ownership, power prices, effects on views and fishing, and potential ecological damage.

Scientific evidence indicates offshore wind farms are relatively safe for marine life and cause less harm than boats and fishing gear. Some studies also suggest the infrastructure can create new habitat for marine life.

However, a general lack of research into offshore wind and marine life has created uncertainty that groups such as Responsible Future Illawarra can exploit.

It has cited statements by Sea Shepherd Australia to argue offshore wind farms damage marine life – however Sea Shepherd said its comments were misrepresented.

3. Australians for Natural Gas

Australians for Natural Gas is a pro-gas group set up by the head of a gas company, which presents itself as a grassroots organisation. Its advertising campaign promotes natural gas as a necessary part of Australia’s fuel mix, and stresses its contribution to jobs and the economy.

The ad campaign implicitly suggests climate action – in this case, a shift to renewable energy – is harmful to the economy, livelihoods and energy security. According to Meta’s Ad Library, these adds have already been seen more than 1.1 million times.

Gas is needed in Australia’s current energy mix. But analysis shows it could be phased out almost entirely if renewable energy and storage was sufficiently increased and business and home electrification continues to rise.

And of course, failing to tackle climate change will cause substantial harm across Australia’s economy.

How to identify misinformation

As the federal election approaches, climate misinformation and disinformation is likely to proliferate further. So how do we distinguish fact from fiction?

One way is through “pre-bunking” – familiarising yourself with common claims made by climate change deniers to fortify yourself against misinformation

Sources such as Skeptical Science offer in-depth analyses of specific claims.

The SIFT method is another valuable tool. It comprises four steps:

• Stop
• Investigate the source
• Find better coverage
• Trace claims, quotes and media to their original sources.

As the threat of climate change grows, a flow of accurate information is vital to garnering public and political support for vital policy change.

Alfie Chadwick, PhD Candidate, Monash Climate Change Communication Research Hub, Monash University and Libby Lester, Professor (Research) and Director, Monash Climate Change Communication Hub, Monash University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.

By comparing what we know about the training datasets of different models, we identified the CommonCrawl dataset of scraped internet pages as the most likely vector where AI models first learned this term.

The scale problem

Finding errors of this sort is not easy. Fixing them may be almost impossible.

One reason is scale. The CommonCrawl dataset, for example, is millions of gigabytes in size. For most researchers outside large tech companies, the computing resources required to work at this scale are inaccessible.

Another reason is a lack of transparency in commercial AI models. OpenAI and many other developers refuse to provide precise details about the training data for their models. Research efforts to reverse engineer some of these datasets have also been stymied by copyright takedowns.

When errors are found, there is no easy fix. Simple keyword filtering could deal with specific terms such as vegetative electron microscopy. However, it would also eliminate legitimate references (such as this article).

More fundamentally, the case raises an unsettling question. How many other nonsensical terms exist in AI systems, waiting to be discovered?

Implications for science and publishing

This “digital fossil” also raises important questions about knowledge integrity as AI-assisted research and writing become more common.

Publishers have responded inconsistently when notified of papers including vegetative electron microscopy. Some have retracted affected papers, while others defended them. Elsevier notably attempted to justify the term’s validity before eventually issuing a correction.

We do not yet know if other such quirks plague large language models, but it is highly likely. Either way, the use of AI systems has already created problems for the peer-review process.

For instance, observers have noted the rise of “tortured phrases” used to evade automated integrity software, such as “counterfeit consciousness” instead of “artificial intelligence”. Additionally, phrases such as “I am an AI language model” have been found in other retracted papers.

Some automatic screening tools such as Problematic Paper Screener now flag vegetative electron microscopy as a warning sign of possible AI-generated content. However, such approaches can only address known errors, not undiscovered ones.

Living with digital fossils

The rise of AI creates opportunities for errors to become permanently embedded in our knowledge systems, through processes no single actor controls. This presents challenges for tech companies, researchers, and publishers alike.

Tech companies must be more transparent about training data and methods. Researchers must find new ways to evaluate information in the face of AI-generated convincing nonsense. Scientific publishers must improve their peer review processes to spot both human and AI-generated errors.

Digital fossils reveal not just the technical challenge of monitoring massive datasets, but the fundamental challenge of maintaining reliable knowledge in systems where errors can become self-perpetuating.

Aaron J. Snoswell, Research Fellow in AI Accountability, Queensland University of Technology; Kevin Witzenberger, Research Fellow, GenAI Lab, Queensland University of Technology, and Rayane El Masri, PhD Candidate, GenAI Lab, Queensland University of Technology

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Return to news.