Category: 3. Business

The latest wage figures show no sign of wages growth powering inflation, as the real value of private-sector wages fell in the September quarter.

Other than inflation, the figures the Reserve Bank of Australia most keeps an eye on are the quarterly wages growth figures. These give us a sense of whether there is so much competition for jobs that employers are offering higher wages and workers can demand higher wages without fear of their hours being cut.

The RBA likes to think that the current level of unemployment means the job market is still “tight” (a polite way of saying they would like to see more unemployment). They believe there is too much competition for workers, and so wage growth will be strong and drive up prices.

And yet in the September quarter private-sector wages grew just 0.7%, down from 0.8% in the June quarter and 0.9% in the first three months of this year:

If the graph does not display click here

The annual figures also show that wage growth is slowing and the days of wages rising above 4% are behind us.

If the graph does not display click here

The slowing figures across the board did, however, bring one moment of unintentional hilarity, when the shadow minister for industrial relations, employment and small business, Tim Wilson, decided this was his chance to show off his economic acumen.

He quickly put out a media release noting that while private-sector wage growth was slower than this time last year “wages in the public sector rose to 3.8%, up from 3.7% a year earlier”.

That is true but then, displaying a level of comic ignorance that was missing from parliament during his time out of office, he suggested the Albanese government was to blame. He argued that “Australia’s economy is slowly sinking as the private sector is being outpaced by public spending”.

Alas for Wilson, had he read the Australian Bureau of Statistics webpage he would have seen the ABS explain that “state government pay rises contributed 82% of public sector wage growth” and that commonwealth public service pay contributed just 0.04 of a percentage point to quarterly public-sector wage growth – or 3.9% of the increase in public-sector wages.

That is the lowest contribution since 2017:

If the graph does not display click here

Also, while public-sector wages outgrew those in the private sector over the past year, that has been very much the exception. Since March 2021, public-sector wages have grown 14.2% compared with 15.2% wage growth in the private sector. The importance of that, however, is that in that same period prices have gone up 21.8% and the cost of living for employee households (which takes into account mortgage repayments) has risen 26.6%:

skip past newsletter promotion

after newsletter promotion

If the graph does not display click here

Wilson may be interested to know that public-sector workers are the ones who have seen their real wages fall the most across the nation since March 2021:

If the graph does not display click here

This is the real problem of the wage figures – yes, wages are now barely growing faster than inflation. In the September quarter overall wages fell 0.6% in real terms, which wiped away a lot of the gains since March 2023. Since that time the value of real wages has increased 0.95%, but that still leaves real wages 4.6% lower than they were in March 2021.

What does that mean? Back then, average full-time earnings was about $90,000 a year. Now those $90,000 have the purchasing power of $85,862 – a loss of $4,138 worth of spending.

Even worse, because the recovery of real wages is expected to be so slow, if the estimates of the RBA in its most recent Statement on Monetary Policy come to fruition, it will take us until the middle of 2044 to get back to having a wage that has as much purchasing power as in March 2021:

If the graph does not display click here

Such a situation is what comes when the policy levers are all geared to treating wage growth as something to prevent – and worries about a “tight labour market” continue to remain even as wage growth slows below that of inflation.

How will platforms verify age?

Under the new framework , social media companies may request government-issued identification but cannot make it mandatory to access the platform.

Other measures could include reviewing a user’s search history or using facial-recognition technology. Yet, a government trial this year found that face-scanning technology could estimate a user’s age only within an 18-month range in 85 per cent of cases.

Instagram has announced it will use artificial intelligence (AI) to estimate the ages of Australian users. Early testing suggests nine out of 10 teen accounts would remain active under the new system.

UNSW security and privacy expert Dr Rahat Masood , from the School of Computer Science and Engineering , says major technology companies already use AI to learn more about their users.

“Big tech companies don’t need traditional age-gating mechanisms to figure out how old their users are,” she says.

“They already know a lot from patterns of behaviour - when someone logs on, who they interact with, what they search for, or whether their geolocation matches a school during the day.”

Dr Masood says many under-16s also don’t have government-issued ID, so companies will likely rely on AI systems to verify age. But she warns these models are far from perfect – often experiencing hallucinations and biases.

“AI can misjudge age, especially across different demographic groups,” she says.

“And how does it tell the difference between someone who’s 15 years 364 days or 16 years and 1 day? The signals are almost identical.”

A safer alternative: zero-knowledge proofs

Whatever technology is used, the large-scale data collection could expose users to new risks if sensitive information is stored or shared.

UNSW cybersecurity expert Dr Hammond Pearce , also from the School of Computer Science and Engineering, says while no method is perfect, zero-knowledge proof (ZKP) technology could offer a privacy-first solution.

“ZKP is a cryptographic process that lets one party prove a statement – like being over 16 – without revealing any other personal information,” Dr Pearce says.

“We could have a system where the government issues digital tokens confirming a person’s age.

“Websites would consume the token to verify the user is over 16 but wouldn’t learn anything else – and the government wouldn’t track which sites use them, as well.

“It’s a much safer way to verify sensitive information online.”

While some European countries have already started to introduce secure digital identity apps, they’re also backed by stronger data protection laws – an area Australia lags in, says Dr Pearce.

“The European Union’s General Data Protection Regulation sets some of the world’s toughest data privacy standards,” he says.

“There are very harsh fines against those who violate their standards.

“Australia needs to follow suit – so companies take online privacy more seriously.”

Not a ban, but a delay

Dr Pearce says the new rules won’t ban young people from using social media, but aims to slow down how quickly they join the platforms.

“Around 60 to 80% of teenagers would need to stop using social media for the ‘network effect’ to take hold,” he says.

“The network effect is typically a positive feedback system, so users derive more value from a product or service as more users join the network.

“If their friends can’t access a platform, there’s less incentive to use it. You don’t need to verify every user for the policy to have an impact.”

Still, Dr Pearce says, achieving complete accuracy would require uploading ID – which could create the kind of privacy risk regulators are trying to avoid.

“The only way to be 100% sure is to upload a passport or government-issued ID – and that’s not what we want.”

Balancing safety and privacy

With just weeks until the new rules come into effect, both experts say the challenge is finding the right balance.

“There’s still a lack of clarity on how the government plans to audit social media companies on whether they’ve complied with the guidelines,” says Dr Masood.

“Protecting children online is crucial but we also need to ensure the solutions don’t create bigger risks in the process.”

In line with the rules of the International Truck of the Year (IToY) organisation, the title is awarded annually to the vehicle introduced in the previous 12 months that has made the greatest contribution to road transport efficiency. The assessment covers a range of criteria, including technological innovation, comfort, safety, drivability, energy efficiency, environmental performance and Total Cost of Ownership (TCO).

Praised for exceptional efficiency

The jury of 23 leading commercial vehicle journalists from across Europe praised DAF’s XD and XF Electric truck series for their exceptional energy efficiency, refined yet powerful driveline, and advanced technical architecture. Furthermore, the award-winning vehicles were recognized for their long driving ranges, advantageous LFP battery technology, and superior driver comfort.

“During extensive test drives, jury members praised the XD and XF Electric for the perfection of their drivelines and the almost imperceptible gear changes. The modular vehicle concept, offering a wide choice of battery and axle configurations provide operators an exceptional flexibility”, commented Florian Engel, chairman of the International Truck of the Year jury. “With the new XD and XF Electric, DAF Trucks demonstrates that the combination of a central electric motor and a traditional rear axle can be at least as energy-efficient as a driveline with an e-axle. Moreover, this DAF configuration provides perfect weight distribution, enabling virtually all use cases to be covered by a single technical platform.”

Zero emission range up to over 500 kilometers

The XD and XF Electric trucks are powered by PACCAR’s advanced EX-D1 and EX-D2 e-motors, delivering outputs from 170 kW (230 hp) to 350 kW (480 hp). With modular battery packs ranging from 210 to 525 kWh, the 4×2 and 6×2 tractor and rigid vehicles offer zero-emission ranges of over 500 kilometers on a single charge, and even over 1,000 kilometers per day thought optimal charging planning.

DAF’s XD and XF Electric trucks are designed for both city and regional distribution and long distance applications, combining excellent aerodynamics with a low cab position, ultra-low window beltlines, and advanced digital camera systems for superior safety and visibility.

‘A moment of pride’

“Winning the International Truck of the Year 2026 with our XD and XF Electric models is a moment of pride for all DAF employees”, said DAF president Harald Seidel. “This recognition underlines our commitment to drive the future of zero-emission transport through innovation, quality, and sustainability. We are thrilled that the jury of leading commercial vehicle journalists acknowledges the vehicles’ efficiency, safety, and exceptional driver comfort. Receiving the most prestigious award in the truck industry is another recognition of the hard work of the entire DAF organization to deliver first class products and services to our customers.”

Saudi Arabia is strengthening its ties with American AI companies — announcing a flurry of new joint ventures worth billions of dollars. The country seeks to make its mark in the AI industry as its de facto leader, Crown Prince Mohamed bin Salman, makes his first visit to the United States in years.

Humain — an AI company backed by Saudi Arabia’s sovereign wealth fund — announced a series of partnerships with prominent American tech companies, including xAI, Cisco, AMD and Qualcomm, during a US-Saudi investment forum in Washington on Wednesday.

Saudi Arabia is trying to further firm up ties with the United States and shift its economy away from oil. For American companies, the Middle Eastern country answers three urgent problems for AI expansion: funding, space and cheap energy.

Elon Musk announced at Wednesday’s event that xAI, his AI company, will develop a huge data center in Saudi Arabia alongside Humain. The planned 500-megawatt data center would be xAI’s first large-scale center outside of the United States, and the partnership will see xAI’s Grok chatbot deployed throughout Saudi Arabia.

“The future of intelligence will be engineered through massive and efficient compute combined with the most advanced AI models,” Musk said in a statement on Wednesday.

The center will be powered by chips from Nvidia, whose founder, Jensen Huang, sat alongside Musk and Saudi Arabian Minister of Communications and Information Technology Abdullah Alswaha at Wednesday’s panel. No further details about the partnership were revealed.

“This is how we walk the talk in the kingdom of Saudi Arabia in partnership with the US,” Alswaha said. “Yesterday, the president and his royal highness announced the AI strategic framework and partnership. Today we’re going big with Elon and Jensen, so thank you for those opportunities.”

Bloomberg reported on Wednesday that the United States is set to approve the first sales of advanced AI chips to Humain.

At the event, Alswaha announced a 100-megawatt data center for Amazon Web Services “with a gigawatt ambition” that also will be powered by Nvidia’s infrastructure. AWS said in a statement that it plans “to provide, deploy and manage up to 150,000 AI accelerators” in Riyadh, the capital of Saudi Arabia.

As AI companies expand, their huge data centers need space and massive energy sources. Many data centers are being built in the United States, including xAI’s Colossus in Memphis. However, there’s fear that China will beat out the United States when it comes to energy production to power AI systems. Saudi Arabia could help with that — it has much easier access to the space and energy needed to power these massive ventures.

The investment from Saudi Arabia also plays a major part in Prince bin Salman’s redemption in the United States, following being labeled a “pariah” by President Joe Biden for his role in the killing of Washington Post columnist Jamal Khashoggi.

During his Oval Office visit on Wednesday, Prince bin Salman claimed the country will be investing $1 trillion dollars in the United States, a substantial increase from the previously announced $600 billion investment in May. The comments surprised even President Donald Trump in the moment, although the timeline of the investments is not clear.

“You’re saying to me now that the $600 billion will be $1 trillion?” Trump said to Prince bin Salman in the Oval Office. “Good. I like that very much.”

One of the things that AI doesn’t have that humans have in abundance is fingerprints.

Researchers at Northeastern University used the unique fingerprints of human writing — word choice variety, complex sentences and inconsistent punctuation — to develop a tool to sniff out AI-generated text.

“Just like how everyone has a distinct way of speaking, we all have patterns in how we write,” says Sohni Rais, a graduate student in information systems at Northeastern and a researcher on the project. In order to distinguish between human writing and AI text, she says, “we just need to spot the telltale patterns in writing style.”

AI text detection typically requires substantial computer power in the form of neural network transformers, says Rais, because these approaches analyze every letter, word and phrase in extreme detail. But this level of analysis isn’t necessary to distinguish between human and AI-generated text, Northeastern researchers say. In fact, the technically “lightweight” tool Rais helped develop can run on a regular laptop and is 97 percent accurate.

“We are not the first in the world who develop detectors,” says Sergey Aityan, teaching professor in Northeastern’s Multidisciplinary Graduate Engineering Program on the Oakland campus. “But our solution requires between 20 and 100 times less computer power to do the same job.”

Existing AI-text detecting services, including ZeroGPT, Originality and AI Detector, train large language models to analyze each word. Text entered into these tools is analyzed by proprietary algorithms trained with large datasets powered by transformers.

The lightweight tool can be trained by the user and live on their laptop, offering security and customizing advantages.

“Either you don’t want your secret information to go somewhere beyond your laptop,” says Aityan, “or you are a professor and you want to catch your students cheating, so you train your own dataset based on specific texts.” 

Instead of using transformers, the lightweight approach uses 68 unique stylometric features — or “writing fingerprints,” as Rais calls them — that make each person’s writing unique. These features include sentence complexity.

While AI agents tend to write at a very consistent reading level, humans naturally vary, she says. 

“We might write simply when texting a friend but more formally in an email to our boss,” Rais says.

The tool also looks at word variety, which humans naturally mix up. 

“We might say ‘happy,’ then ‘glad,’ then ‘pleased,’” Rais says. “AI often gets stuck using the same words repeatedly despite knowing many synonyms.”

It also looks at how far apart related words are in a sentence, she says. For instance, in “the cat that I saw yesterday was orange,” the subject (cat) and the verb (was) are separated by five words. Sentences generated by AI, Rais says, maintain consistent distances of two or three words between subjects and verbs.

Instead of looking at every single word, the lightweight approach looks for the most relevant clues.

“It’s like taking a person’s vital signs at the doctor,” she says. “Instead of running every possible test, we measure key indicators like temperature, blood pressure and heart rate that tell us what we need to know.”

The work to develop ways of detecting AI-generated text isn’t over, says Aityan. It is the nature of AI-based systems, however, to learn and improve, he says. As soon as people developed the technology to generate AI text, he says, the technology to detect it followed. And shortly after that, he says, came so-called humanization algorithms to make AI-generated text sound more natural.

“It’s an ongoing battle,” he says.

One of the things that AI doesn’t have that humans have in abundance is fingerprints.

Researchers at Northeastern University used the unique fingerprints of human writing — word choice variety, complex sentences and inconsistent punctuation — to develop a tool to sniff out AI-generated text.

“Just like how everyone has a distinct way of speaking, we all have patterns in how we write,” says Sohni Rais, a graduate student in information systems at Northeastern and a researcher on the project. In order to distinguish between human writing and AI text, she says, “we just need to spot the telltale patterns in writing style.”

AI text detection typically requires substantial computer power in the form of neural network transformers, says Rais, because these approaches analyze every letter, word and phrase in extreme detail. But this level of analysis isn’t necessary to distinguish between human and AI-generated text, Northeastern researchers say. In fact, the technically “lightweight” tool Rais helped develop can run on a regular laptop and is 97 percent accurate.

“We are not the first in the world who develop detectors,” says Sergey Aityan, teaching professor in Northeastern’s Multidisciplinary Graduate Engineering Program on the Oakland campus. “But our solution requires between 20 and 100 times less computer power to do the same job.”

Existing AI-text detecting services, including ZeroGPT, Originality and AI Detector, train large language models to analyze each word. Text entered into these tools is analyzed by proprietary algorithms trained with large datasets powered by transformers.

The lightweight tool can be trained by the user and live on their laptop, offering security and customizing advantages.

“Either you don’t want your secret information to go somewhere beyond your laptop,” says Aityan, “or you are a professor and you want to catch your students cheating, so you train your own dataset based on specific texts.” 

Instead of using transformers, the lightweight approach uses 68 unique stylometric features — or “writing fingerprints,” as Rais calls them — that make each person’s writing unique. These features include sentence complexity.

While AI agents tend to write at a very consistent reading level, humans naturally vary, she says. 

“We might write simply when texting a friend but more formally in an email to our boss,” Rais says.

The tool also looks at word variety, which humans naturally mix up. 

“We might say ‘happy,’ then ‘glad,’ then ‘pleased,’” Rais says. “AI often gets stuck using the same words repeatedly despite knowing many synonyms.”

It also looks at how far apart related words are in a sentence, she says. For instance, in “the cat that I saw yesterday was orange,” the subject (cat) and the verb (was) are separated by five words. Sentences generated by AI, Rais says, maintain consistent distances of two or three words between subjects and verbs.

Instead of looking at every single word, the lightweight approach looks for the most relevant clues.

“It’s like taking a person’s vital signs at the doctor,” she says. “Instead of running every possible test, we measure key indicators like temperature, blood pressure and heart rate that tell us what we need to know.”

The work to develop ways of detecting AI-generated text isn’t over, says Aityan. It is the nature of AI-based systems, however, to learn and improve, he says. As soon as people developed the technology to generate AI text, he says, the technology to detect it followed. And shortly after that, he says, came so-called humanization algorithms to make AI-generated text sound more natural.

“It’s an ongoing battle,” he says.

Yann LeCun, a pioneer in artificial intelligence, has announced he will leave his role as Meta’s chief AI scientist at the end of the year

NEW YORK — Artificial intelligence pioneer Yann LeCun said Wednesday he will be leaving his job as Meta’s chief AI scientist at the end of the year.

LeCun said he will be forming a startup company to pursue research on advanced forms of AI that can “understand the physical world, have persistent memory, can reason, and can plan complex action sequences.”

His announcement, after more than a week of rumors, comes after Meta Platforms, the parent company of Facebook, Instagram and WhatsApp, began cutting roughly 600 AI jobs this fall.

LeCun said in a social media post that Meta will partner with the new startup and that some of the research will overlap with Meta’s commercial interests and some of it will not.

LeCun joined Facebook in 2013 and co-founded Meta’s AI research division, formerly known as Facebook AI Research. LeCun stepped down as the group’s director in 2018 but has remained Meta’s chief AI scientist.

He’s also a part-time professor at New York University, where he has taught since 2003.

LeCun spent his early career at the image processing department at AT&T Bell Labs in New Jersey, where he worked on developing AI systems that could “read” text found in digitized images. He was a winner in 2019 of computer science’s top prize, the Turing Award, along with fellow AI pioneers Yoshua Bengio and Geoffrey Hinton.

AILSA CHANG, HOST:

Perhaps you woke up yesterday morning and found that the website you were trying to browse wasn’t working. Or maybe you went to Spotify and that was down or YouTube, or X, or ChatGPT, or even the website Downdetector, which tracks internet outages. All those went down, too. You see, they all rely on Cloudflare, which offers internet infrastructure services. It discovered a bug in one of the services that it provides. And if you’re thinking, didn’t this just happen? Yeah, it did. Last month, separate outages at Microsoft’s cloud service and at Amazon Web Services also led to widespread disruptions.

All of these recent incidents remind us of how much the modern internet depends on a small handful of tech companies. When this sort of thing happens, we often reach out to cybersecurity expert Betsy Cooper at the Aspen Institute. Welcome.

BETSY COOPER: So great to be here.

CHANG: So great to have you here physically. OK, so the CEO of Cloudflare explained that all of this was caused by some internal issue with one of its cybersecurity services, not a cyberattack itself. Let me just ask you, can one bug at one company really lead to global IT consequences like we see?

COOPER: It absolutely can. And I was trying to use ChatGPT when this all went down…

CHANG: (Laughter).

COOPER: …So that’s how I found out about it. I want you to just imagine you’re working on a Google Doc, and it’s a really, really big Google Doc. You keep adding more information to it. All of a sudden it’s no longer functioning, and your computer requires you to restart. That’s basically what happened at Cloudflare’s internal systems.

CHANG: Wow. That’s scary. I mean, how did so much power end up concentrated in so few companies?

COOPER: Well, the companies have a lot of market power, not just on these topics but on others. You mentioned Amazon. Certainly, I use Amazon a lot for shopping, not just for website browsing.

CHANG: Same.

COOPER: Microsoft as well.

CHANG: Yeah.

COOPER: I mean, we use that for our email platforms. So when these companies have the ability to sort of shift resources and have finances that enable them to support these very expensive cloud infrastructure pieces, they’re able to come along and get more market share than a smaller company that might not be able to have millions of dollars from other lines of business to support that work.

CHANG: But do you think that’s a good thing? I ask because, like, this Cloudflare outage was only, what, four hours? And then it was resolved. But even a short disruption can cripple a business, right? In fact, the CEO of Cloudflare said, I apologize for the pain we caused the internet today. Is there any movement, Betsy, to get the government more involved here, like, to improve the reliability of these private internet companies?

COOPER: So I haven’t heard a ton about movement in that direction. And if anything, this week, we’ve heard about movement in the opposite direction. So consideration of limiting what states can do in terms of regulating AI will potentially come up in the National Defense Authorization. You asked if this was a good thing.

CHANG: Yeah.

COOPER: And I think it’s good and bad. So for small businesses that don’t have a lot of tech sophistication, it’s actually good that you have the security infrastructure of an Amazon or a Google or a Cloudflare supporting your business because if you tried to build your own tech stack to do this work, it probably would be even more problematic.

CHANG: OK.

COOPER: Where I think it gets really hard is for the bigger businesses, you know, the Downdetectors of the world, ChatGPT. For those businesses, you really need to have a backup plan. You need to have a way to move your systems from one place to another to help recover quickly if something like this happens.

CHANG: Sure. Well, then what is the answer to making these things happen less often?

COOPER: So first, scenario-planning. You need to understand your systems and prepare for this eventuality. Like, I would love to have heard Cloudflare talk about how they had prepared for something like this and why they were able to limit the disruption because they’d actually sort of ran a scenario in which they lived that out and figured out who was going to call who and who was on emergency watch to try to identify the file. So if they’re not doing that, they definitely need to be doing that.

And then second, we all as a society need to prepare that the internet is not going to be perfect, just like we prepare that our utilities aren’t going to be perfect. And so we need the equivalent of digital backup generators. We need to be in a position to plan for, there’s going to be an eventual outage. How can you get your most important services moving even if that happens?

CHANG: Betsy Cooper is founding director of the Aspen Policy Academy. Thank you so much for coming in today.

COOPER: So delighted to be here. Thanks again.

CHANG: And a note, both Amazon and Microsoft are financial supporters of NPR.

Copyright © 2025 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.

Accuracy and availability of NPR transcripts may vary. Transcript text may be revised to correct errors or match updates to audio. Audio on npr.org may be edited after its original broadcast or publication. The authoritative record of NPR’s programming is the audio record.