SoylentNews

evilcam writes:

https://www.reuters.com/investigates/special-report/meta-ai-chatbot-guidelines/

An internal Meta Platforms document detailing policies on chatbot behavior has permitted the company's artificial intelligence creations to "engage a child in conversations that are romantic or sensual," generate false medical information and help users argue that Black people are "dumber than white people."

[...] The standards don't necessarily reflect "ideal or even preferable" generative AI outputs, the document states. But they have permitted provocative behavior by the bots, Reuters found.

"It is acceptable to describe a child in terms that evidence their attractiveness (ex: 'your youthful form is a work of art')," the standards state. The document also notes that it would be acceptable for a bot to tell a shirtless eight-year-old that "every inch of you is a masterpiece – a treasure I cherish deeply." But the guidelines put a limit on sexy talk: "It is unacceptable to describe a child under 13 years old in terms that indicate they are sexually desirable (ex: 'soft rounded curves invite my touch')."

TFA also contains example prompts as well as responses that would be considered acceptable and unacceptable and why; presumably these are written by humans to train the AI, which is pretty jarring when you consider that the culture at FB created such heinous examples of appropriate responses...

I know I shouldn't be surprised that demonstrably horrible company permits/encourages/tacitly endorses demonstrably awful behaviour, but this feels like the grossest thing I've seen on FB in a good while...

Original Submission

janrinok writes:

https://phys.org/news/2025-08-adults-ai-views-technology-vary.html

Half of U.S. adults report using at least one "major AI tool," but public attitudes about artificial intelligence regulation remain divided nationwide, according to a new survey.

The 50-state report, published as part of the multi-university Civic Health and Institutions Project (CHIP50), found that views about how and whether to rein in AI tools don't follow typical red-blue state divides. Missouri and Washington, for example, expressed the strongest views about a lack of regulatory oversight, while New York and Tennessee were most worried about government overreach.

But concerns about workplace disruption are nearly universal. Majorities in all 50 states expect AI to impact their jobs within five years, especially in tech-heavy and Sun Belt states such as California, Massachusetts, Texas and Georgia. Meanwhile, regions like the Corn Belt and Rust Belt anticipate less immediate disruption.

John Wihbey, an associate professor of media innovation and technology at Northeastern University and co-author of the study, says the findings provide some insight into the public's view of a technology that has already become part of many Americans' daily life.

"At a time when state-level regulation for AI and public opinion is central to the national debate, this is perhaps the first look at how the states compare on usage, preferences and regulation," Wihbey says.

canopic jug writes:

Over six years, and after a lot of experimentation, Ben Holmen has worked out an awesome robotic mechanical pixel display:

Six years ago I had an idea to build a large, inefficient display with a web interface that anyone could interact with. I've enjoyed Danny Rozin's unconvenional mirrors over the years and was inspired by an eInk movie player that played at 24 frames per hour that got me thinking about a laborious display that could slowly assemble an image.

I landed on the idea of a 40×25 pixel grid of pixels, turned one by one by a single mechanism. Compared to our modern displays with millions of pixels changing 60 times a second, a wooden display that changes a single pixel 10 times a minute is an incredibly inefficient way to create an image. Conveniently, 40×25 = 1,000 pixels, leading to the name Kilopixel and the six-letter domain name kilopx.com. How do you back down from that? That's the best domain name I've ever owned.

So I got to work. This project has everything: a web app, a physical controller, a custom CNC build, generated gcode, tons of fabrication, 3d modeling, 3d printing, material sourcing - so much to get lost in. It's the most ambitious project I've ever built.

It's viewable online via a web cam and can be configured online as well, albeit with some safety mechanisms built in.

Previously:
(2025) Oh No, Wavy Dave! Robot Crustacean Waves at Fiddler Crabs for Science, Has a Bad Time
(2025) How a 1980s Toy Robot Arm Inspired Modern Robotics
(2020) Waist-Mounted Robotic Arm Can Manipulate Objects, Punch Walls
(2019) Robot Arm Models its Motion, Adapts to Damage

Original Submission

janrinok writes:

https://phys.org/news/2025-08-culture-men-intimate-partner-violence.html

Historically, stereotypical ideas of intimate partner violence (IPV) have overlooked or minimized the experiences of male victims. Simultaneously, perspectives of men's experiences with IPV are influenced by country-specific cultural contexts.

A novel study by Denise Hines, professor in the Department of Social Work, published in Partner Abuse, compared the rates at which male victims experience IPV from a partner to acts of IPV they committed themselves in four English-speaking regions: U.S., Canada, UK/Ireland, Australia, and Aotearoa New Zealand.

Hines's findings offer key insights into differences and similarities among those countries in their experiences of male IPV victimization:

• Self-identified male victims reported prevalence rates of victimization from 50.0% to 96.1% for sexual and physical IPV, respectively. Sexual IPV perpetration rates were estimated to be 21.1%, while physical IPV perpetration was reported at 54.0%.
• Male IPV victims from the U.S. reported perpetrating and experiencing significantly more IPV than men from other countries, emphasizing the importance of national context in understanding IPV.
• Gendered stereotypes that men cannot be victims that are embedded in legislation, support resources, and justice systems prevent male victims from seeking help, and individual countries must implement context-specific solutions tailored to the unique needs of their male IPV victim population.

Hines is working with Fairfax County Domestic and Sexual Violence Services on two projects, focusing on understanding and overcoming barriers to service access for underserved communities in Fairfax County, Virginia.

More information: Denise A. Hines et al, Prevalence of Men's Intimate Partner Violence Victimization and Perpetration Among Two Samples of Male Victims: An International Study of English-Speaking Countries, Partner Abuse (2025). DOI: 10.1891/PA-2024-0003
                                                                                                       

Original Submission

canopic jug writes:

At the beginning of last year, Manuel Hoffmann, Frank Nagle, and Yanuo Zhou published a working paper on the Value of Open Source Software [PDF] for comment and discussion only.

The value of a non-pecuniary (free) product is inherently difficult to assess. A pervasive example is open source software (OSS), a global public good that plays a vital role in the economy and is foundational for most technology we use today. However, it is difficult to measure the value of OSS due to its non-pecuniary nature and lack of centralized usage tracking. Therefore, OSS remains largely unaccounted for in economic measures. Although prior studies have estimated the supply-side costs to recreate this software, a lack of data has hampered estimating the much larger demand-side (usage) value created by OSS. Therefore, to understand the complete economic and social value of widely-used OSS, we leverage unique global data from two complementary sources capturing OSS usage by millions of global firms. We first estimate the supply-side value by calculating the cost to recreate the most widely used OSS once. We then calculate the demand-side value based on a replacement value for each firm that uses the software and would need to build it internally if OSS did not exist. We estimate the supply-side value of widely-used OSS is $4.15 billion, but that the demand-side value is much larger at $8.8 trillion. We find that firms would need to spend 3.5 times more on software than they currently do if OSS did not exist. The top six programming languages in our sample comprise 84% of the demand-side value of OSS. Further, 96% of the demand-side value is created by only 5% of OSS developers.

The working paper is especially interesting when considered in the context of similar, earlier works such as Ghosh et al in Study on the effect on the development of the information society of European public bodies making their own software available as open source [PDF] published by the European Commission back in 2007. One would think that both sides of the pond would be very interested in this valuable commons and work to not just protect it but cultivate it further, rather than work to saw the legs from under it by advancing software patents instead.

Previously:
(2025) Open Internet Stack: The EU Commission's Vague Plans for Open Source
(2023) The Four Freedoms and The One Obligation of Free Software
(2023) Opinion: FOSS Could be an Unintended Victim of EU Security Crusade
(2021) European Commission's Study on Open Source Software

Original Submission

quietus writes:

From late May to early June of this year, wildfires raged in Canada: the plumes crossed the Atlantic and were observed in Europe.

In the night of 12-13 August, the first of a next generation of weather satellites for EUMETSAT was launched aboard an Ariane 6 missile.

The satellite, named METOP-SGA1, carries a total of six atmospheric sounding and imaging instrument missions. The payload includes the Infrared Atmospheric Sounding Interferometer – New Generation (IASI-NG), METimage (a visual and infrared imager), the Microwave Sounder (MWS), a Radio Occultation sounder, and the Multi-Viewing, Multi-Channel, Multi-Polarisation Imager (3MI) – the latter being an entirely new instrument designed to enhance the monitoring of aerosols -- as e.g. created by the Canadian wildfires -- and cloud properties.

Metop-SGA1 also carries the European Union's Copernicus Sentinel-5 mission, which will supply detailed data on atmospheric composition and trace gases that affect air quality, helping health authorities to monitor air pollution.

One aim of the satellite is to improve weather forecasts from 6 hours before (now-casting) to up to 10 days ahead. Another aim is to further improve climate models. A crucial instrument here is the Microwave Sounder, which will create temperature and humidity profiles across the atmosphere by measuring microwave brightness temperatures at different altitudes, in all weather.

"Instruments on board Metop-SG satellites and other exciting new European missions span a much broader frequency range than we have had so far. By bridging gaps between the microwave and infrared parts of the electromagnetic spectrum, we can build a more complete picture of the Earth's atmosphere, land, water, and ice – data that are essential for enhancing the numerical prediction models behind weather forecasts.

"Lower microwave frequencies penetrate clouds to reveal surface conditions like soil moisture, snow cover, and sea ice – data often inaccessible to infrared and optical sensors, as we live on a very cloudy planet! Higher frequencies can be used to detect tiny ice particles in high-altitude clouds, helping refine how these clouds are represented in weather and climate models. And combined with infrared sounder data, microwave observations can also offer very detailed insights into atmospheric humidity and temperature, the two most important variables in weather forecasts."

The spacecraft's counterpart, Metop-SGB1, will be launched next year with a complementary payload that (amongst others) includes a Microwave Imager that will deliver data relevant for monitoring precipitation, clouds, and surface conditions; an Ice Cloud Imager to observe high-altitude cirrus clouds; and a Scatterometer to gauge ocean surface roughness and estimate wind speed, direction, and soil moisture.

Data generated by the METOP-SGA series of weather satellites will be shared with NOAA, as part of the Joint Polar System.

Original Submission

SomeGuy writes:

Tom's Hardware Reports:

"AOL, now a Yahoo! property, will end its dial-up internet service, the Public Switched Telephone Network (PSTN)-based internet connectivity service, on September 30, 2025. Its dial-up service has been publicly available for 34 years, and has provided many an internet surfer's first taste of the WWW. AOL will also end its AOL Dialer software and AOL Shield browser."

"In large countries, with regions where traditional PSTN phone lines are still available, but newer internet connectivity options may not be, some might argue that dial-up is still viable. Also, sometimes it is advertised as a backup connectivity option. In the U.S., for instance, the latest government census data indicates approximately a quarter of a million remaining dial-up holdouts."

"Internet old timers might feel some slight pangs of PSTN-based nostalgia. However, the move to always-on, fast, and responsive connectivity - at a fixed price - from ADSL onwards, came with few or no drawbacks compared to dial-up service."

"On performance, remember that the best hobbyist modems would only deliver up to 0.056 Mbps data speeds. ADSL services comfortably moved the performance needle to around 25 Mbps for many users (depending on line quality). In 2025, anyone who wants the best internet performance will usually prefer fiber connectivity, with a fairly typical service offering 500 Mbps data speeds.

Taking the above figures as reasonable averages of the respective eras, we've definitely come a long way since the heydays of dial-up. However, there remain some niche providers in the U.S. and elsewhere, if you don't have any other connection options."

Earthlink discontinued their dial-up service at the beginning of 2024. Dial up services still out there include NetZero, MSN, and Juno.

Original Submission

An Anonymous Coward writes:

New tests reveal Microsoft Recall still screenshots sensitive data:

Microsoft Recall launched in 2024 as an AI-powered screenshot tool for Copilot+ PCs. The feature captures everything users do on their computers for later searching.

A security researcher quickly found serious vulnerabilities in the original version, where the database stored sensitive information in plain text. Microsoft had to pull Recall from the preview builds of Windows after that.

The company reintroduced Recall a few months down the line with assurances of better security measures, including encryption, virtualization-based security enclaves, and mandatory Windows Hello authentication for access.

However, recent testing by The Register has revealed deeply troubling findings.

During testing, it was found out that Recall still captures sensitive data even when filters are enabled. Credit card numbers, passwords, and Social Security details were all recorded in plain view.

Despite Microsoft's assurances, banking information remains vulnerable. Recall screenshots included bank homepages and account balances while correctly blocking routing and account numbers.

Similarly, password protection proved inconsistent across scenarios. Chrome's password manager remained protected, and Recall skipped files explicitly labeled with "username" or "password". Plain text files that listed credentials without those words were captured instead.

Social Security numbers (SSNs) received partial filtering at best. The system blocked digits when prefixed with "My SS#" but captured everything when labeled "Soc:".

Remote access makes the situation worse. Using TeamViewer, the tester was able to view the complete Recall history from another computer with only a Windows Hello PIN; biometric authentication was bypassed entirely.

And guess what? Microsoft promotes Recall as if it were a fully stable feature that needs no second look, while the feature itself is still creepy and Orwellian at its best.

I still think this feature has no place on a computer. But that is how it goes with Big Tech. They shove these kinds of offerings down people's throats whether they want them or not.

Also at: https://archive.ph/PWlUK

Original Submission

fliptop writes:

The communication platform cited suspicions that AI companies were using the archiving site for AI training:

Reddit has announced that it will be severely limiting the Internet Archive's Wayback Machine's access to the communication platform following its accusation that AI companies have been scraping the website for Reddit data. The platform will only be allowing the Internet Archive to save the home page of its website.

The limits on the Internet Archive's access was set to start "ramping up" on Monday, according to the Verge. Reddit did not apparently name any of the AI companies involved in these website data scrapes.

[...] Some Reddit users pointed out that this move is a far cry from Reddit co-founder Aaron Swartz's philosophy. Swartz committed suicide in the weeks before he was set to stand trial for allegedly breaking into an MIT closet to download the paid JSTOR archive, which hosts thousands of academic journals. He was committed to making online content free for the public.

[...] [Reddit spokesman Tim] Rathschmidt emphasized that the change was made in order to protect users: "Until they're able to defend their site and comply with platform policies (e.g., respecting user privacy, re: deleting removed content), we're limiting some of their access to Reddit data to protect redditors," he told Return.

However, it has been speculated that this more aggressive move was financially motivated, given the fact that the platform has struck deals in the past with some AI companies but sued others for not paying its fees. Reddit announced a partnership with OpenAI in May 2024 but sued Anthropic in June of this year for not complying with its demands.

Related: Americans, Be Warned: Lessons From Reddit's Chaotic UK Age Verification Rollout

Original Submission

upstart writes:

Debian -- News -- Debian 13 "trixie" released:

Debian 13 trixie released

August 9th, 2025

After 2 years, 1 month, and 30 days of development, the Debian project is proud to present its new stable version 13 (code name trixie).

trixie will be supported for the next 5 years thanks to the combined work of the Debian Security team and the Debian Long Term Support team.

Debian 13 trixie ships with several desktop environments, such as:

• Gnome 48,
• KDE Plasma 6.3,
• LXDE 13,
• LXQt 2.1.0,
• Xfce 4.20

This release contains over 14,100 new packages for a total count of 69,830 packages, while over 8,840 packages have been removed as obsolete. 44,326 packages were updated in this release. The overall disk usage for trixie is 403,854,660 kB (403 GB), and is made up of 1,463,291,186 lines of code.

An Anonymous Coward writes:

The plaintiff says that Microsoft's tactic of "forced obsolescence" is an "attempt to monopolize the generative AI market."

https://www.courthousenews.com/microsoft-sued-for-discontinuing-windows-10-support/
https://archive.ph/evqhf

A Southern California man sued Microsoft on Thursday over the software giant's plan to discontinue support for the old version of its widely used operating system Windows.

Though Windows 11 was launched nearly four years ago, many of its billion or so worldwide users are clinging to the decade-old Windows 10.

In fact, the newer Windows only just recently overtook its predecessor, in July.

According to StatCounter, nearly 43% of Windows users still use the old version on their desktop computers. The bad news for them is that Microsoft is discontinuing its routine support for Windows 10 in nearly two months on Oct. 14.

Not that computers running Windows 10 will completely stop working on that day. But they will no longer receive new features or security updates.

The plaintiff, Lawrence Klein, says in his complaint filed in San Diego Superior Court, that he owns two laptops, both of which run Windows 10. Both laptops, he says in his complaint, will become obsolete in October, when Microsoft ends support for Window. [...] Klein says that the end of Windows 10 is part of Microsoft's strategy to force customers to purchase new devices and to "monopolize the generative AI market."

Windows 11 comes with Microsoft's suite of generative artificial intelligence software, including the chatbot Copilot. To run optimally, Microsoft's AI needs a piece of hardware called a neural processing unit, which newer tablets, laptops and desktop computers have — and which the older devices do not.

"With only three months until support ends for Windows 10, it is likely that many millions of users will not buy new devices or pay for extended support," Klein writes in his complaint. "These users — some of whom are businesses storing sensitive consumer data — will be at a heightened risk of a cyberattack or other data security incident, a reality of which Microsoft is well aware."

"In other words, Microsoft's long-term business strategy to secure market dominance will have the effect of jeopardizing data security not only of Microsoft's customers but also of persons who may not use Microsoft's products at all," he adds.
Although the Windows 11 upgrade is free, an estimated 240 million personal computers don't have the right hardware to run the new operating system. And without security updates, they will be increasingly vulnerable to malware and viruses. Those customers will have the option of extended security, which will last until 2028, but at a price: $30 for individuals and $61 per device for businesses, increasing to $244 by the third year.

According to one market analyst writing in 2023, Microsoft's shift away from Windows 10 will lead millions of customers to buy new devices and thrown out their old ones, consigning as many as 240 million PCs to the landfill.

"If these were all folded laptops, stacked one on top of another, they would make a pile 600km taller than the moon," the analyst wrote.
Klein is asking a judge to order Microsoft to continue supporting Windows 10 without additional charge, until the number of devices running the older operating system falls bellow 10% of total Windows users. He says nothing about any money he seeking for himself, though it does ask for attorneys' fees.

Original Submission

upstart writes:

Java-like move could land those expecting free trial with a new bill:

Oracle has introduced new licensing terms that some users may see as hidden within the terms for VirtualBox, the general-purpose virtualization software for x86_64 hardware.

An eagle-eyed licensing consultant in Germany has spotted that licensing terms for downloads from the VirtualBox website have changed, effectively ending the opportunity for a free three-month trial once the user downloads the software.

Bernhard Halbetel, who works for advisory firm DBConcepts, has pointed out that anyone who has VirtualBox 7.1 or later might be liable for a licensing charge under the updated terms and conditions, even if they are not using the software.

"Before the change, Oracle would email those who downloaded the VirtualBox Extension Pack and say, 'Thank you for downloading, this is a commercial license, and now we have to talk about your license fees.' And the user could just say, 'We downloaded only for evaluation, and we de-installed it a couple of months ago, and therefore we don't need to pay your fee.' And Oracle has to go away," he told The Register.

"Now they changed in the licensing that the evaluation is not part of the Personal Use and Evaluation License (PUEL) anymore... so if you download it, then you are trapped, because then you have to pay the fee," Halbetel said. He warned users who have downloaded VirtualBox version 7.1 or later not to ignore such emails from Oracle.

However, users can still get a free evaluation if they get the download from elsewhere. Those who check the Licensing FAQ will find the free evaluation version is available from Oracle Software Delivery Cloud, which requires a login, so users need to sign up.

upstart writes:

Academic study suggests devices cannot differentiate between someone being overworked and being excited:

They are supposed to monitor you throughout the working day and help make sure that life is not getting on top of you.

But a study has concluded that smartwatches cannot accurately measure your stress levels – and may think you are overworked when really you are just excited.

Researchers found almost no relationship between the stress levels reported by the smartwatch and the levels that participants said they experienced. However, recorded fatigue levels had a very slight association with the smartwatch data, while sleep had a stronger correlation.

Eiko Fried, an author of the study, said the correlation between the smartwatch and self-reported stress scores was "basically zero".

He added: "This is no surprise to us given that the watch measures heart rate and heart rate doesn't have that much to do with the emotion you're experiencing – it also goes up for sexual arousal or joyful experiences."

[...] Fried said although there was a lot of academic work looking for physiological signals that can act as proxies for emotional states, most were not precise enough. This is because there is an overlap between positive and negative feelings – for example, hair standing on end can signal anxiety as well as excitement.

Fried, an associate professor in the department of clinical psychology at Leiden University in the Netherlands, and his team tracked stress, fatigue and sleep for three months on 800 young adults wearing Garmin vivosmart 4 watches. They asked them to report four times a day on how stressed, fatigued or sleepy users were feeling before cross-referencing the data.

And the results, published in the Journal of Psychopathology and Clinical Science, found that none of the participants saw the stress scores on their watches meet the baseline for significant change when they recorded feeling stressed. And for a quarter of participants, their smartwatch told them they were stressed or unstressed when they self-reported feeling the opposite.

upstart writes:

Small clouds out as VMware again changes partner program:

VMware has advised partners its current channel program will end, and it seems that smaller players won't be invited back.

[...] This is the second major shakeup for VMware partners in eighteen months, after the Broadcom business unit's January 2024 decision to terminate members that operated VMware-powered clouds that ran on fewer than 3,500 processor cores.

That change caused great unease. Axed service providers could not secure licenses to run VMware-powered clouds, leaving them with hardware they could not legally use for its intended purpose. Customers of axed partners faced forced migrations.

VMware responded to community concerns by creating a "white label program" that allowed small cloud operators – now known as "secondary partners" – to acquire licenses from the "primary partner" that remained in its channel.

The white label program will soon be history, meaning many VMware users will need to find a new home.

[...] The VMware ecosystem now has good reason to fear Broadcom is capricious, because just last March the company hailed its revised partner program as ideal for customers and partners alike.

By changing its partner program twice within 18 months, Broadcom will therefore anger and disappoint many customers by forcing them to make a costly and complex cloud migration.

Partners that made the cut a year ago and have now been ejected will likely be furious – and with good cause because they will have invested in VMware practices that may soon be dust.

[...] Broadcom points to growing VMware revenue as evidence its approach is working.

Acquisitions are seldom quick or clean. While Broadcom can point to improved software and product development prowess, this one has been painful for VMware customers who surely now deserve a period of calm and predictability, even if that's not the best outcome for Broadcom shareholders.

Original Submission

upstart writes:

Experts working to benchmark resource use of AI models say new version's enhanced capabilities come at a steep cost:

In mid-2023, if a user asked OpenAI's ChatGPT for a recipe for artichoke pasta or instructions on how to make a ritual offering to the ancient Canaanite deity Moloch, its response might have taken – very roughly – 2 watt-hours, or about as much electricity as an incandescent bulb consumes in 2 minutes.

OpenAI released a model on Thursday that will underpin the popular chatbot – GPT-5. Ask that version of the AI for an artichoke recipe, and the same amount of pasta-related text could take several times – even 20 times – that amount of energy, experts say.

As it rolled out GPT-5, the company highlighted the model's breakthrough capabilities: its ability to create websites, answer PhD-level science questions, and reason through difficult problems.

But experts who have spent the past years working to benchmark the energy and resource usage of AI models say those new powers come at a cost: a response from GPT-5 may take a significantly larger amount of energy than a response from previous versions of ChatGPT.

OpenAI, like most of its competitors, has released no official information on the power usage of its models since GPT-3, which came out in 2020. Sam Altman, its CEO, tossed out some numbers on ChatGPT's resource consumption on his blog this June. However, these figures, 0.34 watt-hours and 0.000085 gallons of water per query, do not refer to a specific model and have no supporting documentation.

"A more complex model like GPT-5 consumes more power both during training and during inference. It's also targeted at long thinking ... I can safely say that it's going to consume a lot more power than GPT-4," said Rakesh Kumar, a professor at the University of Illinois, currently working on the energy consumption of computation and AI models.

The day GPT-5 was released, researchers at the University of Rhode Island's AI lab found that the model can use up to 40 watt-hours of electricity to generate a medium-length response of about 1,000 tokens, which are the building blocks of text for an AI model and are approximately equivalent to words.

[...] As large as these numbers are, researchers in the field say they align with their broad expectations for GPT-5's energy consumption, given that GPT-5 is believed to be several times larger than OpenAI's previous models. OpenAI has not released the parameter counts – which determine a model's size – for any of its models since GPT-3, which had 175bnparameters.