SoylentNews

Freeman writes:

https://arstechnica.com/ai/2025/10/ai-powered-search-engines-rely-on-less-popular-sources-researchers-find/

Since last year's disastrous rollout of Google's AI Overviews, the world at large has been aware of how AI-powered search results can differ wildly from the traditional list of links search engines have generated for decades. Now, new research helps quantify that difference, showing that AI search engines tend to cite less popular websites and ones that wouldn't even appear in the Top 100 links listed in an "organic" Google search.

In the pre-print paper "Characterizing Web Search in The Age of Generative AI," researchers from Ruhr University in Bochum, Germany, and the Max Planck Institute for Software Systems compared traditional link results from Google's search engine to its AI Overviews and Gemini-2.5-Flash. The researchers also looked at GPT-4o's web search mode and the separate "GPT-4o with Search Tool," which resorts to searching the web only when the LLM decides it needs information found outside its own pre-trained data.
[...]
Overall, the sources cited in results from the generative search tools tended to be from sites that were less popular than those that appeared in the top 10 of a traditional search, as measured by the domain-tracker Tranco. Sources cited by the AI engines were more likely than those linked in traditional Google searches to fall outside both the top 1,000 and top 1,000,000 domains tracked by Tranco. Gemini search in particular showed a tendency to cite unpopular domains, with the median source falling outside Tranco's top 1,000 across all results.
[...]
For search terms pulled from Google's list of Trending Queries for September 15, the researchers found GPT-4o with Search Tool often responded with messages along the lines of "could you please provide more information" rather than actually searching the web for up-to-date information.

While the researchers didn't determine whether AI-based search engines were overall "better" or "worse" than traditional search engine links, they did urge future research on "new evaluation methods that jointly consider source diversity, conceptual coverage, and synthesis behavior in generative search systems."

Original Submission

An Anonymous Coward writes:

https://9to5linux.com/fedora-linux-43-officially-released-now-available-for-download

This release is powered by the latest and greatest Linux 6.17 kernel series and features both GNOME 49 and KDE Plasma 6.4 desktop environments.

The Fedora Project officially released Fedora Linux 43 today as the latest stable version of this Red Hat-sponsored distribution, shipping with some of the latest and greatest GNU/Linux technologies.

Highlights of Fedora Linux 43 include the latest and greatest Linux 6.17 kernel series, the latest and greatest GNOME 49 desktop environment series for the Fedora Workstation edition, which is now Wayland-only, as well as the KDE Plasma 6.4.5 desktop environment on the Fedora KDE Plasma Desktop edition.

Fedora Linux 43 also brings the Anaconda WebUI installer by default to more Fedora Spins, support for the COLRv1 format in the Noto Color Emoji fonts, support for the Hare programming language, a default Monospace fallback font, and DNF 5 by default on the Anaconda installer for RPM package installation.

Among other changes, Fedora 43 introduces a 2GB boot partition, automated onboarding to Packit release automation for new packages, automatic updates by default in Fedora Kinoite, zstd-compressed initrd by default, package-specific RPM macros for build flags, and a rewrite of Greenboot written in Rust.

This new Fedora Linux release also enforces the use of GPT partition tables for all UEFI-based Fedora installations for 64-bit systems, which removes support for installing Fedora in UEFI mode on MBR-partitioned disks. AArch64 and RISC-V systems remain unaffected.

Under the hood, Fedora 43 features an up-to-date toolchain and components consisting of GCC 15.2, GNU Binutils 2.45, GNU C Library 2.42, GDB 17.1, LLVM 21, Golang 1.25, Perl 5.42, RPM 6.0, Python 3.14, PostgreSQL 18, Ruby on Rails 8.0, Dovecot 2.4, MySQL 8.4, Tomcat 10.1, Apache Maven 4, Haskell GHC 9.8, and Idris 2.

https://fedoraproject.org/
https://docs.fedoraproject.org/en-US/quick-docs/upgrading-fedora-offline/

Original Submission

"The definition of insanity is doing the same thing over and over again and expecting different results." :- misattributed to Einstein

"We cannot solve our problems with the same thinking we used when we created them." :- Einstein

There is a lot in this Meta but it is necessary to have certain aspects of the site's operation explained in detail so that subsequent elements make sense and are understandable by everyone. The initial lessons from the Trial of Flagging by Journal Owners appear later in this Meta.

Permanent Banning

Banning someone from the site is a serious decision which is why it is rarely considered. It has always been recognised that the act of banning someone is never going to be easy to enforce. Some may wonder why banning is even considered at all, and the explanation is relatively simple. Some acts – in this case doxxing – can have serious repercussions and kolie has described elsewhere those potentially applicable under US law an in particular to the state of Oregon.

The rules exist to ensure that we maintain a viable site where people are free to discuss the topics presented in an adversarial yet friendly atmosphere. If the site's rules are not enforced then they are meaningless and, over time, they will be ignored. We are very tolerant of minor infractions but at some point it is necessary to remind somebody of the reason for the rule and that is often all that is needed. The next level usually involves moderation, possibly with an Admin-To-User message warning that the user might receive a temporary ban if (s)he continues. If temporary bans do not work then, in extreme cases, it is necessary to employ a permanent ban. Permanent bans require the approval of the Board.

Admin-to-User messages can only be used to communicate from staff to account holders; the only way to communicate with Anonymous Cowards is directly via a comment.

Doxxing

During the last few days some have challenged the definition of doxxing. In particular, they have argued that the addresses given are obviously fake and therefore are not doxxing. The rules are quite clear. It isn't possible for staff to recognise every address as being genuine or fake and so it is always assumed to be genuine and treated as such.

Kolie's amusing and robust counter can be seen here.

Sock-Puppets and Multiple Accounts

Each person is allowed to have one account which gives him the right to vote on the site. They are owned by a community member and they are not transferable nor can they be shared accounts. Additional accounts can be created providing that they are notified to, and agreed by, the Administration and are required to fulfill a specific function e.g upstart and Arthur T Knackerbracket (story submission bots), Acfriendly (journal to facilitate AC participation in front page stories) etc. These additional bots do not have voting rights nor should they receive moderation points.

Fake accounts are those accounts created to use the site often by persons not intending to participate in discussions. They are usually created for advertising purposes and might also be created by commercial organisations. As most people never even see them they cause little problem other than take up an account identity. However, occasionally they engage in activities that are not aligned to the site's purpose and they are disabled. Accounts created entirely in, or sometimes using, foreign languages are disabled as a matter of routine. Some of these have been associated with material that is illegal under US Federal or State laws.

Sock Puppets are accounts that are created usually with the intention of giving the user an unfair advantage with regards to voting or moderation abuse. They are sometimes intended to give the user an alternative account to use when their primary account would attract a ban, or to use when their primary account has already be given a temporary ban. Very often the sock puppet account is employed to positively moderate inflammatory or abusive posts made Anonymous Cowards, thus preventing the community from controlling such material by selecting a reasonable viewing level, while leaving the sock puppet account apparently innocent of any wrongdoing.

Historically, some users have created multiple sock-puppets as were used increasingly during the "Sock-Puppet-War" between 2018 and 2021. Each user employed the sock-puppets in an attempt to prevent the other from expressing his or her personal view and with the further hope that their opponent might be banned from the site. One person has created hundreds of sock-puppets. The names of many of these sock puppets contain doxxing material or express unacceptable statements.

I have been watching specific sock-puppets for some time to try to understand their purpose. As recently as earlier this week I disabled 6 such accounts. Their creator is known. There are as yet others that I am still analysing.

Real-World Politics

In the USA in particular, but elsewhere too, the political situation has become very polarised. There is much hatred between opposing political factions which has sometimes resulted in violence or other physical and verbal abuse. This site is NOT the place to continue to express your dislike of other community members who do not share your own political views. Demanding that someone be banned or continually punished for having a particular political view is abuse and it will be treated as such by the staff.

Everyone in our community has the same rights to express their opinion, and any attempt to prevent one of them from doing so is unacceptable. If you find yourself having to name a person in a comment it is often the sign that you are intending that comment as a personal attack. If a topic or journal is intended to discuss a political viewpoint it is entirely correct to do so, but that does not include personal attacks against other community members.

Journals

Journals are for account holders to discuss any topic that is legal under US and State laws, but which would not be considered for front page use. The topics do not have to be written to suit everybody in the community. They do not have to meet with the approval of individual community members who have no right to demand that the journal owner stops writing such journals or take other action to intentionally disrupt the subsequent journal discussion.

...And if you have managed to get this far I hope that what I have written will now make more sense than it might have done before and you will now understand its relevance:

Flagging Trial

The removal of non-account Anonymous Cowards from the main pages has made discussion far more acceptable to many people. Unfortunately, it has not had the same effect in the journals which a minority of ACs have been using to disrupt the discussions and abuse the journal owners and other community members. As a result, fewer people are using the journals to introduce their own discussions, and fewer people are participating in journal discussions.

Several journal owners requested that we investigate ways of controlling the abuse. It was apparent that such control would be a significant task for staff with the current software and data. The site has always had a means of removing illegal or unacceptable content from display. From the very first days of the site there has been a facility to delete comments from the database. However, the method involved hard deletes (permanent deletions from the database) but that left the child comments also inaccessible. Soft deleting (flagging) was adopted in 2024 as a far better solution. The use of the flagging is different from the community's perception than the previous system because:

• It is immediately apparent that flagging has taken place. Previously comments just 'disappeared' and were irretrievable.
• The community needs to know that the system is not being abused, which could be provided by having increased visibility of the processes involved
• Such visibility raises several issues – why has a comment been flagged? who would do the flagging? and how would it be managed?

After discussions with some journal owners they agreed to assist in a trial in which journal owners themselves would be able to exercise some control over abuse and/or disruption in their journals. There were 3 journal owners initially and others participated as their journals appeared.

There is one permanently banned account – aristarchus. Even before his ban we have over several years tried various methods including moderation, arranging for him to rejoin the community with some restrictions, and deletion of his comments. This is not new and goes back to the very early days of the site. In 2014 he was already abusing some of the same people that he abuses today. His complaints about blocked IP addresses and censorship go back to at least 2016. This alone indicates that the blocked IP addresses are unrelated to any other function and are automatic within the Rehash software. The site rules state that technical means can be employed to remove such comments and that now implies flagging.

Identifying his posts was initially marred by the occasional mis-identification. Where they were brought to my attention they were corrected and apologies made – publicly and privately. Since the start of 2025 the amount of data available to us has increased in its nature, quantity and accuracy. It is far more reliable today than it was. Nevertheless, there is no automatic flagging and a person remains the final decision maker based on the originator and the contents of the comment in its entirety.

Findings and Recommended Actions

• Journal owners are reluctant to use the flagging mechanism for perfectly understandable reasons. They would prefer an inclusive, community-based discussion. They do not want the abuse and disruption and flagging provides them with the means to control such occurrences should they wish to use it. Action: Consider starting all journals set to Logged-in Users only as default. Journal owners should still have the opportunity to open the discussion wider if they wish.
• There is no reason to remove the facility from journals for those journal owners who might now, or in the future, wish to use flagging to control abuse in their journal. Action: Leave the facility in situ. We may also need it for future trials.
• Some readers still find the reduced banner impairs their ability to read a discussion. Action: Investigate still further whether it is possible for the banner be reduced further in size – perhaps just to the comment number in a smaller font? It is recognised that the fragility of parts of Rehash might make this very difficult to achieve.
• The management of flagging will require additional data to be recorded with each flagged item. For example, if a comment contains doxxing it should record the comment and also set a flag to prevent it ever being released. There may be additional requirements as the existing software is enhanced. Action: Keep as is for the moment but be aware that changes will be required.
• The management of flagging will either require additional staff for it to be maintainable over a period of time, or significant additional software to assist in the management task. Action: If no additional manpower is available the next best option is to make the site Logged-in Users only. Reverting to a previous state (i.e. relying on basic moderation) will only result in the same outcome as it did previously.
• User requests for a flagging to be reviewed (not simply viewed) must be by the person who made the original comment. Action: How to do this for ACs is not yet identified. Otherwise the system can be easily defeated by numerous unjustified requests for reviews by miscellaneous people..
• The decision to flag a comment can for the moment only be made by a person. Despite the process being far more reliable now than it initially was it is still below the level that would result in an automatic system being viable.
• A clear policy that is acceptable to the community must be provided to state clearly when and how flagging is permissible. Action: A policy must be written with community consultation to fulfill this requirement.

Your comments are invited. ACs will have the opportunity to make comments in a journal. While AC views and opinions are welcome any abuse in that journal will be treated appropriately

upstart writes:

Westinghouse is claiming a nuclear deal would see $80B of new reactors:

On Tuesday, Westinghouse announced that it had reached an agreement with the Trump administration that would purportedly see $80 billion of new nuclear reactors built in the US. And the government indicated that it had finalized plans for a collaboration of GE Vernova and Hitachi to build additional reactors. Unfortunately, there are roughly zero details about the deal at the moment.

The agreements were apparently negotiated during President Trump's trip to Japan. An announcement of those agreements indicates that "Japan and various Japanese companies" would invest "up to" $332 billion for energy infrastructure. This specifically mentioned Westinghouse, GE Vernova, and Hitachi. This promises the construction of both large AP1000 reactors and small modular nuclear reactors. The announcement then goes on to indicate that many other companies would also get a slice of that "up to $332 billion," many for basic grid infrastructure.

So the total amount devoted to nuclear reactors is not specified in the announcement or anywhere else. As of the publication time, the Department of Energy has no information on the deal; Hitachi, GE Vernova, and the Hitachi/GE Vernova collaboration websites are also silent on it.

Meanwhile, Westinghouse claims that it will be involved in the construction of "at least $80 billion of new reactors," a mix of AP1000 and AP300 (each named for the MW of capacity of the reactor/generator combination). The company claims that doing so will "reinvigorate the nuclear power industrial base."

That's going to take some work. As of now, there are zero nuclear reactors under construction, and the last two that were completed were enough to bankrupt Westinghouse. (It's now co-owned by Cameco, a nuclear fuel supplier, and Brookfield Asset Management.) The Financial Times reports that one of Westinghouse's owners thinks that the $80 billion should be enough for eight reactors, but would only finance five if they cost as much as the AP1000s previously built in the US. The FT also reports that the US government would share in any profits and a stake in the company if the deal goes forward.

One of the big challenges these deals will face, however, is achieving profitability. According to the Department of Energy's latest evaluation, nuclear power is the second-most expensive source of electricity in the US, behind offshore wind, and the cost of offshore wind has fallen in recent years. Finances aren't the only risk to this deal. None of the designs for small modular reactors developed by any of these companies has currently been approved by the Nuclear Regulatory Commission.

Original Submission

Freeman writes:

https://arstechnica.com/tech-policy/2025/10/senators-move-to-keep-big-techs-creepy-companion-bots-away-from-kids/

The US will weigh a ban on children's access to companion bots, as two senators announced bipartisan legislation Tuesday that would criminalize making chatbots that encourage harms like suicidal ideation or engage kids in sexually explicit chats.

At a press conference, Josh Hawley (R-Mo.) and Richard Blumenthal (D-Conn.) introduced the GUARD Act, joined by grieving parents holding up photos of their children lost after engaging with chatbots.
[...]
Failing to block a minor from engaging with chatbots that are stoking harmful conduct—such as exposing minors to sexual chats or encouraging "suicide, non-suicidal self-injury, or imminent physical or sexual violence"—could trigger fines of up to $100,000, Time reported. (That's perhaps small to a Big Tech firm, but notably higher than the $100 maximum payout that one mourning parent suggested she was offered.)
[...]
It covers any AI chatbot that "provides adaptive, human-like responses to user inputs" and "is designed to encourage or facilitate the simulation of interpersonal or emotional interaction, friendship, companionship, or therapeutic communication," Time reported.
[...]
"In their race to the bottom, AI companies are pushing treacherous chatbots at kids and looking away when their products cause sexual abuse, or coerce them into self-harm or suicide," Blumenthal told NBC News. "Our legislation imposes strict safeguards against exploitative or manipulative AI, backed by tough enforcement with criminal and civil penalties."

Hawley agreed with Garcia that the AI industry must align with America's morals and values, telling NBC News that "AI chatbots pose a serious threat to our kids.

"More than 70 percent of American children are now using these AI products," Hawley said.
[...]
The tech industry has already voiced opposition. On Tuesday, Chamber of Progress, a Big Tech trade group, criticized the law as taking a "heavy-handed approach" to child safety. The group's vice president of US policy and government relations, K.J. Bagchi, said that "we all want to keep kids safe, but the answer is balance, not bans.

"It's better to focus on transparency when kids chat with AI, curbs on manipulative design, and reporting when sensitive issues arise," Bagchi said.

However, several organizations dedicated to child safety online, including the Young People's Alliance, the Tech Justice Law Project, and the Institute for Families and Technology, cheered senators' announcement Tuesday. The GUARD Act, these groups told Time, is just "one part of a national movement to protect children and teens from the dangers of companion chatbots."
[...]
During Tuesday's press conference, Blumenthal noted that the chatbot ban bill was just one initiative of many that he and Hawley intend to raise to heighten scrutiny on AI firms.

Original Submission

upstart writes:

Some portions of the OS are still stuck on light:

Windows 11 launched way back in October 2021 and has become Microsoft's must-have OS thanks to the impending end-of-life for Windows 10. After all that time, there are still significant portions of the OS that don't do dark mode. However, Redmond is making significant progress, bringing a couple of key dialog boxes into compliance.

As of Windows Insider Build 26120.6972 (beta channel) and 26220.6972 (dev channel) or higher, you can now get the Run and Folder Options dialog boxes to appear in dark mode. However, to make them dark, you need to use the ViVeTool, a utility that enables hidden features.

To enable dark mode in the first place, navigate to Settings-Personalization-Colors and select Dark from the "Choose your mode" menu.

Now, download ViVeTool if you don't have the utility already. Create a folder called C:\vive and unzip the ViVeTool's contents to it.

Launch an elevated command prompt by searching for cmd, right-clicking on it, and selecting "Run as administrator."

*SPOILER* (click to show) *SPOILER* (click to hide)

In the command prompt terminal that appears, change your directory to C:\vive.

cd \vive

Then enter the following command, which will enable dark mode for the Run Dialog, Folder Options, and also for File progress windows and "are you sure" popups that appear when you try to delete or rename certain files or folders. The latter two dark mode instances were added as hidden features back in August, but they still aren't part of Windows 11's default setup.

vivetool /enable /

id:57857165,57994323,48433719,49453572,58383338,59270880,59203365

After you've entered the command, reboot your computer. You should now be able to see dark mode when you hit Windows + R to open the Run dialog box and in Explorer when you select Options from the ... menu.

You'll also see dark mode in action when you're copying files and watching the progress bar.

And you'll see dark mode when you try to rename or delete a folder that requires a higher level of permissions.

But no, you still won't see dark mode when you right click on a file or folder and select Properties.

And you don't see it when you attempt to rename something you do have permission to access but which requires a confirmation. For example, when I tried to rename cmd.exe to cmd.ex, I got a dialog box warning me about the dangers of changing a file extension, but it was in white.

It's clear that some of the Windows features that don't obey dark mode are legacy apps that Microsoft doesn't really want you to use anymore - all of Control Panel has a white background for this reason. However, we expect Microsoft to roll out more dark mode features to File Explorer and to make those features work in the production builds of Windows in the near future, because the company has no plans to replace its tried-and-trusted file management utility.

[Ed. note: I'd be interested if anyone is familiar with that Vivetool utility and what your impression is]

Original Submission

Freeman writes:

https://arstechnica.com/space/2025/10/25-years-one-website-iss-in-real-time-captures-quarter-century-on-space-station/

With the milestone just days away, you are likely to hear this week that there has now been a continuous human presence on the International Space Station (ISS) for the past 25 years. But what does that quarter of a century actually encompass?

[...]

Fortunately, the astronauts and cosmonauts on the space station have devoted some of their work time and a lot of their free time to taking photos, filming videos, and calling down to Earth. Much of that data has been made available to the public, but in separate repositories, with no real way to correlate or connect it with the timeline on which it was all created.

That is, not until now. Two NASA contractors, working only during their off hours, have built a portal into all of those resources to uniquely represent the 25-year history of ISS occupancy.

ISS in Real Time, by Ben Feist and David Charney, went live on Monday (October 27), ahead of the November 2 anniversary.

Original Submission

upstart writes:

ASML launches revolutionary lithography scanner for advanced 3D chip packaging:

Last week, ASML introduced the Twinscan XT:260 lithography scanner, the industry's first scanner that has been designed from the ground up for advanced 3D packaging, marking a new era in fab tools.

Advanced packaging technologies like TSMC's Chip-on-Wafer-on-Substrate (CoWoS) are crucially important to achieve the performance scaling necessary to develop artificial intelligence and to evolve supercomputers.

Advanced packaging relies on deposition, etching, lithography, and metrology/inspection tools to make sophisticated chips. But while using these front-end tools is efficient for many steps, they are overengineered for some, and insufficient for others.

"In line with our plans to support our customers in the 3D integration space, we shipped ASML's first product serving Advanced Packaging, the Twinscan XT:260, an i-line scanner offering up to 4x productivity compared to existing solutions.", ASML posted in its Q3 2025 financial results.

ASML's Twinscan XT:260 is an i-line (365 nm) step-and-scan lithography system that processes 300-mm wafers and weds the precision of previous-generation front-end lithography tools with the productivity and flexibility of back-end tools. TSMC claims this provides four times higher productivity compared to 'competing steppers' used for advanced packaging technologies, such as Canon's FPA-5520iV. ASML never named the exact competing product, but Canon's FPA-5520iV is a good bet.

The key advantage of the tool compared to some of the existing machines used for advanced packaging is that it supports a high-dose exposures (340 mJ is mentioned, though it is usually tunable) and a 52 mm × 66 mm image field, enabling the tool to process up to 3,432 mm^2 interposers (4X EUV reticle size) without field stitching, which reduces complexity and speeds up the production cycle. For the sake of truth, it should be noted that Canon's FPA-5520iV LF2 Option supports a 52 mm × 68 mm image field, but this is a stepper, not a scanner.

The system delivers 400 nm resolution, a 35 nm overlay, and offers a large depth of focus (11 µm at 1 µm CD) to enable accurate patterning for redistribution layers (RDLs) through-silicon vias (TSVs), and hybrid bonding structures used by modern packaging methods to integrate multi-chiplet designs. The unit also boasts 775 µm through-silicon alignment capabilities to make it particularly suited for bonded or non-planar wafers, which are common in 3D stacking.

The Twinscan XT:260 relies on ASML's dual-stage platform, so it can expose one wafer while simultaneously aligning the next, which significantly increases its performance. Speaking of performance, the machine can process up to 270 wafers per hour (at a 340 mJ dose) and handle thick (0.775 mm – 1.7 mm) or warped (1 mm) wafers.

With a 400 nm resolution, 35 nm overlay, and the ability to handle thick or warped wafers (up to 1.7 mm), the XT:260 is optimized for technologies Intel's Foveros, TSMC's CoWoS and System-on-Integrated-Chips (SoIC), as well as other high-density die-stacking or interposer technologies which require precise alignment through silicon.

Freeman writes:

https://arstechnica.com/cars/2025/10/an-autonomous-car-for-consumers-lucid-says-its-happening/

Is it possible to be a CEO in 2025 and not catch a case of AI fever? The latest company to catch this particular cold is Lucid, the Saudi-backed electric vehicle startup. Today, it announced a new collaboration with Nvidia to use the latter's hardware and software, with the aim of creating an autonomous vehicle for consumers. Oh, and the AI will apparently design Lucid's production lines.

Formed by refugees from Tesla who saw a chance to improve on their past work, Lucid has already built the most efficient EV on sale in North America.
[...]
"We've already set the benchmark in core EV attributes with proprietary technology that results in unmatched range, efficiency, space, performance, and handling," said interim CEO Marc Winterhoff. "Now, we're taking the next step by combining cutting-edge AI with Lucid's engineering excellence to deliver the smartest and safest autonomous vehicles on the road. Partnering with Nvidia, we're proud to continue powering American innovation leadership in the global quest for autonomous mobility," Winterhoff said.
[...]
Car buyers are starting to cotton on to driver assists like General Motors' Super Cruise, which about 40 percent of customers choose to pay for after the three-year free trial ends, and Lucid must be hoping that offering a far more advanced system, which won't require the human to pay any attention while it is engaged, will help it earn plenty of money.
[...]
Nvidia's industrial platform will let Lucid create its production lines digitally first before committing them to actual hardware. "By modeling autonomous systems, Lucid can optimize robot path planning, improve safety, and shorten commissioning time," Lucid said.

Original Submission

The Australian Government wants AI to pay for copyright fees in move that may be more about getting a piece of the billions invested in AI. ARIA chief executive Annabelle Herd of the Copyright and AI Reference Group (CAIRG) has called the recommendation for a text and data mining exception "a radical change" that has been "put forward with very little evidence".

"Artificial Intelligence presents significant opportunities for Australia and our economy, however it's important that Australian creatives benefit from these opportunities too," Attorney-General Michelle Rowland said.

"Australian creatives are not only world class, but they are also the lifeblood of Australian culture, and we must ensure the right legal protections are in place."

[...] It is a difficult space for governments to regulate as they balance embracing the promised economic boons of AI without cumbersome red tape while also pitching guardrails.

In the lead-up to Labor's economic reform roundtable in August, the Productivity Commission urged against heavy-handed regulation of AI, warning it could smother opportunities.

Among its recommendations was a text and data mining exception – a call that sparked furore.

But Ms Rowland vowed the government would not "weaken copyright protections when it comes to AI".

"The tech industry and the creative sector must now come together and find sensible and workable solutions to support innovation while ensuring creators are compensated," she said.

"The government will support these next steps through the renewed focus tasked to the Copyright and AI reference group."

Original Submission

Booga1 writes:

OpenAI has acquired Software Applications Incorporated (SAI), perhaps best known for the core team that produced what became Shortcuts on Apple platforms. More recently, the team has been working on Sky, a context-aware AI interface layer on top of macOS. The financial terms of the acquisition have not been publicly disclosed.

"AI progress isn't only about advancing intelligence—it's about unlocking it through interfaces that understand context, adapt to your intent, and work seamlessly," an OpenAI rep wrote in the company's blog post about the acquisition. The post goes on to specify that OpenAI plans to "bring Sky's deep macOS integration and product craft into ChatGPT, and all members of the team will join OpenAI."

...Sky, which leverages Apple APIs and accessibility features to provide context about what's on screen to a large language model; the LLM takes plain language user commands and executes them across multiple applications. At its best, the tool aimed to be a bit like Shortcuts, but with no setup, generating workflows on the fly based on user prompts.

It bears some resemblance to features of Atlas, the ChatGPT-driven web browser that OpenAI launched earlier this week, and this acquisition piles on even more evidence that OpenAI has ambitions beyond a question-and-answer chatbot.

OpenAI can use the SAI team's knowledge of the macOS platform to develop new ways for ChatGPT not just to make suggestions about, but to agentically work directly on users' macOS environments.

Original Submission

upstart writes:

AI arms dealer relies on Taiwanese advanced packaging plants for top-specced GPUs:

US manufacturing of Nvidia GPUs is underway and CEO Jensen Huang is celebrating the first Blackwell wafer to come out of TSMC's Arizona chip factory. However, to be part of a complete product, those chips may need to visit Taiwan.

Nvidia first announced plans to produce chips at Fab21 just six months ago..

Speaking during an event in Phoenix on Friday, Huang lauded TSMC's manufacturing prowess while pandering to US President Donald Trump's America First agenda.

"This is the vision of President Trump of reindustrialization — to bring back manufacturing to America, to create jobs, of course, but also this is the single most vital manufacturing industry and the most important technology industry in the world," he said.

But while the silicon may be homegrown, Nvidia remains reliant on Taiwanese packaging plants to turn those wafers into its most powerful and highest-demand GPUs.

Modern GPUs are composed of multiple compute and memory dies. The company's Blackwell family of datacenter chips feature two reticle-sized compute dies along with eight stacks of HBM3e memory, all stitched together using TSMC's CoWoS packaging tech.

Up to this point, all of TSMC's packaging facilities have been located in Taiwan. Amkor, an outsourced semiconductor assembly and test services (OSAT) provider, is working on building an advanced packaging plant in the US capable of stitching together silicon dies using TSMC's chip-on-wafer-on-substrate (CoWoS) tech. But until it's done – expected in 2027 or 2028 – the next stop for Nvidia's wafers will likely be Taiwan.

During TSMC's Q3 earnings call last week, CEO C.C. Wei confirmed the Amkor plan was moving forward, but the site was only now breaking ground.

upstart writes:

Plus spy helping spy: Typhoons teaming up:

Security researchers now say more Chinese crews - likely including Salt Typhoon - than previously believed exploited a critical Microsoft SharePoint vulnerability, and used the flaw to target government agencies, telecommunications providers, a university, and a finance company across multiple continents.

Threat intel analysts at Broadcom-owned Symantec and Carbon Black uncovered additional victims and malware tools the intruders used, and published those and other details about the attacks in a Wednesday report.

In July, Microsoft patched the so-called ToolShell vulnerability (CVE-2025-53770), a critical remote code execution bug in on-premises SharePoint servers. But before Redmond fixed the flaw, Chinese attackers found and exploited it as a zero-day, compromising more than 400 organizations, including the US Energy Department.

Trend Micro's research team says they've uncovered additional evidence of China-aligned groups, specifically Salt Typhoon and its Beijing botnet-building brethren Flax Typhoon, collaborating in "what looks like a single cyber campaign at first sight."

In these attacks, Salt Typhoon (aka Earth Estries, FamousSparrow) performs the initial break-in, then hands the compromised org over to Flax Typhoon (aka Earth Naga).

"This phenomenon, which we have termed 'Premier Pass,' represents a new level of coordination in cyber campaigns, particularly among China-aligned APT actors," the Trend researchers said.

At the time, Microsoft attributed the break-ins to three China-based groups. These included two government-backed groups: Linen Typhoon (aka Emissary Panda, APT27), which typically steals intellectual property, and Violet Typhoon (aka Zirconium, Judgment Panda, APT31), which focuses on espionage and targets former government and military personnel and other high-value individuals.

Microsoft also accused a suspected China-based criminal org, Storm-2603, of exploiting the bug to infect victims with Warlock ransomware.

It now appears other Beijing crews – including Salt Typhoon, which famously hacked America's major telecommunications firms and stole information belonging to nearly every American – also joined in the attacks.

Original Submission

An Anonymous Coward writes:

In a study published last month https://www.nature.com/articles/s41562-025-02297-0 researchers analyze internal sentence representation for both humans and LLMs. It turns out that humans and LLMs use similar tree structures. Quote from their conclusions: "The results also add to the literature showing that the human brain and LLM, albeit fundamentally different in terms of the implementation, can have aligned internal representations of language."

Originally seen on techxplore https://techxplore.com/news/2025-10-humans-llms-sentences-similarly.html:

A growing number of behavioral science and psychology studies have thus started comparing the performance of humans to those of LLMs on specific tasks, in the hope of shedding new light on the cognitive processes involved in the encoding and decoding of language. As humans and LLMs are inherently different, however, designing tasks that realistically probe how both represent language can be challenging.

Researchers at Zhejiang University have recently designed a new task for studying sentence representation and tested both LLMs and humans on it. Their results, published in Nature Human Behavior, show that when asked to shorten a sentence, humans and LLMs tend to delete the same words, hinting at commonalities in their representation of sentences.

"Understanding how sentences are represented in the human brain, as well as in large language models (LLMs), poses a substantial challenge for cognitive science," wrote Wei Liu, Ming Xiang, and Nai Ding in their paper. "We develop a one-shot learning task to investigate whether humans and LLMs encode tree-structured constituents within sentences."

[...] nterestingly, the researchers' findings suggest that the internal sentence representations of LLMs are aligned with linguistics theory. In the task they designed, both humans and ChatGPT tended to delete full constituents (i.e., coherent grammatical units) as opposed to random word sequences. Moreover, the word strings they deleted appeared to vary based on the language they were completing the task in (i.e., Chinese or English), following language-specific rules.

"The results cannot be explained by models that rely only on word properties and word positions," wrote the authors. "Crucially, based on word strings deleted by either humans or LLMs, the underlying constituency tree structure can be successfully reconstructed."

Overall, the team's results suggest that when processing language, both humans and LLMs are guided by latent syntactic representations, specifically tree-structured sentence representations. Future studies could build on this recent work to further investigate the language representation patterns of LLMs and humans, either using adapted versions of the team's word deletion task or entirely new paradigms.

Journal Reference: Liu, W., Xiang, M. & Ding, N. Active use of latent tree-structured sentence representation in humans and large language models. Nat Hum Behav (2025). https://doi.org/10.1038/s41562-025-02297-0

Original Submission

"Anonymouse" writes:

Trump Eyes Government Control of Quantum Computing Firms With Intel-Like Deals

Donald Trump is eyeing taking equity stakes in quantum computing firms in exchange for federal funding, The Wall Street Journal reported.

At least five companies are weighing whether allowing the government to become a shareholder would be worth it to snag funding that the Trump administration has "earmarked for promising technology companies," sources familiar with the potential deals told the WSJ.

IonQ, Rigetti Computing, and D-Wave Quantum are currently in talks with the government over potential funding agreements, with minimum awards of $10 million each, some sources said. Quantum Computing Inc. and Atom Computing are reportedly "considering similar arrangements," as are other companies in the sector, which is viewed as critical for scientific advancements and next-generation technologies.

No deals have been completed yet, sources said, and terms could change as quantum-computing firms weigh the potential risks of government influence over their operations.

Original Submission