An anonymous reader quotes a report from 404 Media: Anthropic, the company that made one of the most popular AI writing assistants in the world, requires job applicants to agree that they won't use an AI assistant to help write their application. "While we encourage people to use AI systems during their role to help them work faster and more effectively, please do not use AI assistants during the application process," the applications say. "We want to understand your personal interest in Anthropic without mediation through an AI system, and we also want to evaluate your non-AI-assisted communication skills. Please indicate 'Yes' if you have read and agree." Anthropic released Claude, an AI assistant that's especially good at conversational writing, in 2023. This question is in almost all of Anthropic's nearly 150 currently-listed roles, but is not in some technical roles, like mobile product designer. It's included in everything from software engineer roles to finance, communications, and sales jobs at the company. The field was spotted by Simon Willison, an open source developer. The question shows Anthropic trying to get around a problem it's helping create: people relying so heavily on AI assistants that they struggle to form opinions of their own. It's also a moot question, as Anthropic and its competitors have created AI models so indistinguishable from human speech as to be nearly undetectable.

Read more of this story at Slashdot.

A new update is being rolled out to Windows 11 insiders (Build 26120.3073) that introduces a Copilot button in Microsoft Paint. PCWorld reports: Clicking the Copilot button will expand a drop-down menu with all the generative AI features: Cocreator and Image Creator (AI art based on what you've drawn or text prompts), Generative Erase (AI removal of unwanted stuff from images), and Remove Background. Note that these generative AI features have been in Microsoft Paint for some time, but this quick-access Copilot button is a nice time-saver and productivity booster if you use them a lot.

Read more of this story at Slashdot.

NetChoice has filed (PDF) its 10th lawsuit challenging state internet regulations, this time opposing Maryland's Age-Appropriate Design Code Act. The Verge's Lauren Feiner reports: NetChoice has become one of the fiercest -- and most successful -- opponents of age verification, moderation, and design code laws, all of which would put new obligations on tech platforms and change how users experience the internet. [...] NetChoice's latest suit opposes the Maryland Age-Appropriate Design Code Act, a rule that echoes a California law of a similar name. In the California litigation, NetChoice notched a partial win in the Ninth Circuit Court of Appeals, which upheld the district court's decision to block a part of the law requiring platforms to file reports about their services' impact on kids. (It sent another part of the law back to the lower court for further review.) A similar provision in Maryland's law is at the center of NetChoice's complaint. The group says that Maryland's reporting requirement lets regulators subjectively determine the "best interests of children," inviting "discriminatory enforcement." The reporting requirement on tech companies essentially mandates them "to disparage their services and opine on far-ranging and ill-defined harms that could purportedly arise from their services' 'design' and use of information," NetChoice alleges. NetChoice points out that both California and Maryland have passed separate online privacy laws, which NetChoice Litigation Center director Chris Marchese says shows that "lawmakers know how to write laws to protect online privacy when what they want to do is protect online privacy." Supporters of the Maryland law say legislators learned from California's challenges and "optimized" their law to avoid questions about speech, according to Tech Policy Press. In a blog analyzing Maryland's approach, Future of Privacy Forum points out that the state made some significant changes from California's version -- such as avoiding an "express obligationâ to determine users' ages and defining the "best interests of children." The NetChoice challenge will test how well those changes can hold up to First Amendment scrutiny. NetChoice has consistently maintained that even well-intentioned attempts to protect kids online are likely to backfire. Though the Maryland law does not explicitly require the use of specific age verification tools, Marchese says it essentially leaves tech platforms with a no-win decision: collect more data on users to determine their ages and create varied user experiences or cater to the lowest common denominator and self-censor lawful content that might be considered inappropriate for its youngest users. And similar to its arguments in other cases, Marchese worries that collecting more data to identify users as minors could create a "honey pot" of kids' information, creating a different problem in attempting to solve another.

Read more of this story at Slashdot.

An anonymous reader quotes a report from 404 Media: The Air Force Research Laboratory (AFRL), whose tagline is "Win the Fight," has paid more than a hundred thousand dollars to a company that is providing generative AI services to other parts of the Department of Defense. But the AFRL refused to say what exactly the point of the research was, and provided page after page of entirely blacked out, redacted documents in response to a Freedom of Information Act (FOIA) request from 404 Media related to the contract. [...] "Ask Sage: Generative AI Acquisition Accelerator," a December 2023 procurement record reads, with no additional information on the intended use case. The Air Force paid $109,490 to Ask Sage, the record says. Ask Sage is a company focused on providing generative AI to the government. In September the company announced that the Army was implementing Ask Sage's tools. In October it achieved "IL5" authorization, a DoD term for the necessary steps to protect unclassified information to a certain standard. 404 Media made an account on the Ask Sage website. After logging in, the site presents a list of the models available through Ask Sage. Essentially, they include every major model made by well-known AI companies and open source ones. Open AI's GPT-4o and DALL-E-3; Anthropic's Claude 3.5; and Google's Gemini are all included. The company also recently added the Chinese-developed DeepSeek R1, but includes a disclaimer. "WARNING. DO NOT USE THIS MODEL WITH SENSITIVE DATA. THIS MODEL IS BIASED, WITH TIES TO THE CCP [Chinese Communist Party]," it reads. Ask Sage is a way for government employees to access and use AI models in a more secure way. But only some of the models in the tool are listed by Ask Sage as being "compliant" with or "capable" of handling sensitive data. [...] [T]he Air Force declined to provide any real specifics on what it paid Ask Sage for. 404 Media requested all procurement records related to the Ask Sage contract. Instead, the Air Force provided a 19 page presentation which seemingly would have explained the purpose of the test, while redacting 18 of the pages. The only available page said "Ask Sage, Inc. will explore the utilization of Ask Sage by acquisition Airmen with the DAF for Innovative Defense-Related Dual Purpose Technologies relating to the mission of exploring LLMs for DAF use while exploring anticipated benefits, clearly define needed solution adaptations, and define clear milestones and acceptance criteria for Phase II efforts."

Read more of this story at Slashdot.

A century after quantum mechanics revolutionized physics, scientists still cannot agree on how the theory fundamentally works, despite its tremendous success in explaining natural phenomena and enabling modern technologies. The theory's central puzzle remains unresolved: the way quantum systems are described mathematically differs from what scientists observe when measuring them. This has led to competing interpretations about whether quantum states represent physical reality or are merely tools for calculating probabilities. As researchers debate these foundational questions, quantum mechanics has enabled breakthroughs in particle physics, chemistry, and computing. It accurately predicts phenomena from the behavior of atoms to the properties of the Higgs boson, and underlies technologies like quantum computers and ultra-precise measurement devices. The field's inability to reach consensus on its foundations hasn't hindered its practical applications. Scientists continue to develop new quantum technologies even as they grapple with deep questions about measurement, locality, and the nature of reality that have persisted since Einstein and Bohr's famous debates in the 1920s and 1930s.

Read more of this story at Slashdot.

U.S. President Donald Trump signed an executive order on Monday ordering the U.S. Treasury and Commerce Departments to create a sovereign wealth fund and said it may purchase TikTok. From a report: "We're going to stand this thing up within the next 12 months. We're going to monetize the asset side of the U.S. balance sheet for the American people," Treasury Secretary Scott Bessent told reporters. "There'll be a combination of liquid assets, assets that we have in this country as we work to bring them out for the American people." Trump had previously floated such a government investment vehicle as a presidential candidate, saying it could fund "great national endeavors" like infrastructure projects such as highways and airports, manufacturing, and medical research. Details on how exactly the fund would operate and be financed were not immediately available, but Trump previously said it could be funded by "tariffs and other intelligent things." Typically such funds rely on a country's budget surplus to make investments, but the U.S. operates at a deficit.

Read more of this story at Slashdot.

AI startup Anthropic has demonstrated a new technique to prevent users from eliciting harmful content from its models, as leading tech groups including Microsoft and Meta race to find ways that protect against dangers posed by the cutting-edge technology. From a report: In a paper released on Monday, the San Francisco-based startup outlined a new system called "constitutional classifiers." It is a model that acts as a protective layer on top of large language models such as the one that powers Anthropic's Claude chatbot, which can monitor both inputs and outputs for harmful content. The development by Anthropic, which is in talks to raise $2 billion at a $60 billion valuation, comes amid growing industry concern over "jailbreaking" -- attempts to manipulate AI models into generating illegal or dangerous information, such as producing instructions to build chemical weapons. Other companies are also racing to deploy measures to protect against the practice, in moves that could help them avoid regulatory scrutiny while convincing businesses to adopt AI models safely. Microsoft introduced "prompt shields" last March, while Meta introduced a prompt guard model in July last year, which researchers swiftly found ways to bypass but have since been fixed.

Read more of this story at Slashdot.

Cloudflare, a major web infrastructure company, will now track and verify the authenticity of images across its network through Content Credentials, a digital signature system that documents an image's origin and editing history. The technology, developed by Adobe's Content Authenticity Initiative, embeds metadata showing who created an image, when it was taken, and any subsequent modifications - including those made by AI tools. Major news organizations including the BBC, Wall Street Journal and New York Times have already adopted the system. The feature is available immediately through a single toggle in Cloudflare Images settings. Users can verify an image's authenticity through Adobe's web tool or Chrome extension.

Read more of this story at Slashdot.

The exponential rise in microplastic pollution over the past 50 years may be reflected in increasing contamination in human brains, according to a new study. From a report: It found a rising trend in micro- and nanoplastics in brain tissue from dozens of postmortems carried out between 1997 and 2024. The researchers also found the tiny particles in liver and kidney samples. The human body is widely contaminated by microplastics. They have also been found in blood, semen, breast milk, placentas and bone marrow. The impact on human health is largely unknown, but they have been linked to strokes and heart attacks. The scientists also found that the concentration of microplastics was about six times higher in brain samples from people who had dementia. However, the damage dementia causes in the brain would be expected to increase concentrations, the researchers said, meaning no causal link should be assumed. "Given the exponentially rising environmental presence of micro- and nanoplastics, this data compels a much larger effort to understand whether they have a role in neurological disorders or other human health effects," said the researchers, who were led by Prof Matthew Campen at the University of New Mexico in the US.

Read more of this story at Slashdot.

OpenAI has filed an application with the U.S. Patent and Trademark Office to trademark hardware products under its brand name, signaling potential expansion into consumer devices. The filing covers AI-assisted headsets, smart wearables and humanoid robots with communication capabilities. CEO Sam Altman told The Elect on Sunday that OpenAI plans to develop AI hardware through multiple partnerships, though he estimated prototypes would take "several years" to complete.

Read more of this story at Slashdot.

U.S. Representative Zoe Lofgren has introduced a bill that would allow courts to block access to foreign websites primarily engaged in copyright infringement. The Foreign Anti-Digital Piracy Act would enable rightsholders to obtain injunctions requiring large Internet service providers and DNS resolvers to block access to pirate sites. The bill marks a shift from previous site-blocking proposals, notably including DNS providers like Google and Cloudflare with annual revenues above $100 million. Motion Picture Association CEO Charles Rivkin backed the measure, while consumer group Public Knowledge criticized it as "censorious." The legislation requires court review and due process before any blocking orders can be issued. Sites would have 30 days to contest preliminary orders.

Read more of this story at Slashdot.

The Beatles' final song "Now and Then," featuring John Lennon's AI-restored vocals from a 1970s demo, has won the Grammy for Best Rock Performance. Paul McCartney and Ringo Starr completed the track in 2023 using machine learning to isolate Lennon's voice from the original piano recording.

Read more of this story at Slashdot.

Meta's investment in virtual and augmented reality is set to exceed $100 billion this year as CEO Mark Zuckerberg declares 2025 a "defining year" for its smart glasses ambitions. The company invested $19.9 billion in its Reality Labs division last year, according to its annual report, bringing total spending on VR and AR development to over $80 billion since 2014. The unit, which develops Ray-Ban Meta smart glasses and Quest VR headsets, sold 1 million pairs of glasses in 2024 but continues to post losses, according to Financial Times.

Read more of this story at Slashdot.

The blog OMG Ubuntu reports: Ubuntu's key developers have agreed to switch to Matrix as the primary platform for real-time development communications involving the distro. From March, Matrix will replace IRC as the place where critical Ubuntu development conversations, requests, meetings, and other vital chatter must take place... Only the current #ubuntu-devel and #ubuntu-release Libera IRC channels are moving to Matrix, but other Ubuntu development-related channels can choose to move — officially, given some projects were using Matrix over IRC already. As a result, any major requests to/of the key Ubuntu development teams with privileged access can only be actioned if requests are made on Matrix. Canonical-employed Ubuntu developers will be expected to be present on Matrix during working hours... The aim is to streamline organisation, speed up decision making, ensure key developers are reliably reachable, and avoid discussions and conversations from fragmenting across multiple platforms... It's hoped that in picking one platform as the 'chosen one' the split in where the distro's development discourse takes place can be reduced and greater transparency in how and when decisions are made restored. IRC remains popular with many Ubuntu developers but its old-school, lo-fi nature is said to be off-putting to newer contributors. They're used to richer real-time chat platforms with more features (like discussion history, search, offline messaging, etc). It's felt this is why many newer developers employed by Canonical prefer to discuss and message through the company's internal Mattermost instance — which isn't publicly accessible. Many Ubuntu teams, flavours, and community chats already take place on Matrix... "End-users aren't directly affected, of course," they point out. But an earlier post on the same blog notes that Matrix "is increasingly ubiquitous in open-source circles. GNOME uses it, KDE embraces it, Linux Mint migrated last year, Mozilla a few years before, and it's already widely used by Ubuntu community members and developers." IRC remains unmatched in many areas but is, rightly or wrongly, viewed as an antiquated communication platform. IRC clients aren't pretty or plentiful, the syntax is obtuse, and support for 'modern' comforts like media sending, read receipts, etc., is lacking.To newer, younger contributors IRC could feel ancient or cumbersome to learn. Though many of IRC's real and perceived shortcomings are surmountable with workarounds, clients, bots, scripts, and so on, support for those varies between channels, clients, servers, and user configurations. Unlike IRC, which is a centralised protocol relying on individual servers, Matrix is federated. It lets users on different servers to communicate without friction. Plus, Matrix features encryption, message history, media support, and so, meeting modern expectations.

Read more of this story at Slashdot.

Just because you get fired doesn't mean that your pull requests are automatically closed. Dallin was in the middle of reviewing a PR by Steve when the email came out announcing that Steve no longer worked at the company.

Let's take a look at that PR, and maybe we can see why.

$originalUndrawn = DecimalHelper::toDecimal($party->limit)->sub(DecimalHelper::toDecimal($party->drawn));

This is the original code, which represents operations on investments. An investment is represented by a note, and belongs to one or more partys. The amount that can be drawn is set by a limit, which can belong to either the party or the note.

What our developer was tasked with doing was allow a note to have no limit. This means changing all the places where the note's limit is checked. So this is what they submitted:

if ($note->limit == null) { $originalUndrawn = DecimalHelper::toDecimal($party->limit)->sub(DecimalHelper::toDecimal($party->drawn)); } else { $originalUndrawn = DecimalHelper::toDecimal($party->limit)->sub(DecimalHelper::toDecimal($party->drawn)); }

You'll note here that the note limit isn't part of calculating the party limits, so both branches do the same thing. And then there's the deeper question of "is a null really the best way to represent this?" especially given that elsewhere in the code they have an "unlimited" flag that disables limit checking.

Now, Steve wasn't let go only for their code- they were just a miserable co-worker who liked to pick fights in pull request comments. So the real highlight of Steve's dismissal was that Dallin got to have a meaningful discussion about the best way to make this change with the rest of the team, and Steve didn't have a chance to disrupt it.

[Advertisement] Keep the plebs out of prod. Restrict NuGet feed privileges with ProGet. Learn more.

To capitalize on the AI boom, many crypto miners "have begun to repurpose parts of their operations into data centers," reports Reuters, "given they already have most of the infrastructure" (including landing and "significant" power resources...) Toronto-based bitcoin miner Bitfarms has enlisted two consultants to explore how it can transform some of its facilities to meet the growing demand for artificial intelligence data centers, it said on Friday... Earlier this month, Riot Platforms launched a review of the potential AI and computing uses for parts of its facility in Navarro County, Texas.

Read more of this story at Slashdot.

Android and Google Play have billions of users, Google wrote in its security blog this week. "However, like any flourishing ecosystem, it also attracts its share of bad actors... That's why every year, we continue to invest in more ways to protect our community." Google's tactics include industry-wide alliances, stronger privacy policies, and "AI-powered threat detection." "As a result, we prevented 2.36 million policy-violating apps from being published on Google Play and banned more than 158,000 bad developer accounts that attempted to publish harmful apps. " To keep out bad actors, we have always used a combination of human security experts and the latest threat-detection technology. In 2024, we used Google's advanced AI to improve our systems' ability to proactively identify malware, enabling us to detect and block bad apps more effectively. It also helps us streamline review processes for developers with a proven track record of policy compliance. Today, over 92% of our human reviews for harmful apps are AI-assisted, allowing us to take quicker and more accurate action to help prevent harmful apps from becoming available on Google Play. That's enabled us to stop more bad apps than ever from reaching users through the Play Store, protecting users from harmful or malicious apps before they can cause any damage. Starting in 2024 Google also "required apps to be more transparent about how they handle user information by launching new developer requirements and a new 'Data deletion' option for apps that support user accounts and data collection.... We're also constantly working to improve the safety of apps on Play at scale, such as with the Google Play SDK Index. This tool offers insights and data to help developers make more informed decisions about the safety of an SDK." And once an app is installed, "Google Play Protect, Android's built-in security protection, helps to shield their Android device by continuously scanning for malicious app behavior." Google Play Protect automatically scans every app on Android devices with Google Play Services, no matter the download source. This built-in protection, enabled by default, provides crucial security against malware and unwanted software. Google Play Protect scans more than 200 billion apps daily and performs real-time scanning at the code-level on novel apps to combat emerging and hidden threats, like polymorphic malware. In 2024, Google Play Protect's real-time scanning identified more than 13 million new malicious apps from outside Google Play [based on Google Play Protect 2024 internal data]... According to our research, more than 95 percent of app installations from major malware families that exploit sensitive permissions highly correlated to financial fraud came from Internet-sideloading sources like web browsers, messaging apps, or file managers. To help users stay protected when browsing the web, Chrome will now display a reminder notification to re-enable Google Play Protect if it has been turned off... Scammers may manipulate users into disabling Play Protect during calls to download malicious Internet-sideloaded apps. To prevent this, the Play Protect app scanning toggle is now temporarily disabled during phone or video calls... Google Play Protect's enhanced fraud protection pilot analyzes and automatically blocks the installation of apps that may use sensitive permissions frequently abused for financial fraud when the user attempts to install the app from an Internet-sideloading source (web browsers, messaging apps, or file managers). Building on the success of our initial pilot in partnership with the Cyber Security Agency of Singapore (CSA), additional enhanced fraud protection pilots are now active in nine regions — Brazil, Hong Kong, India, Kenya, Nigeria, Philippines, South Africa, Thailand, and Vietnam. In 2024, Google Play Protect's enhanced fraud protection pilots have shielded 10 million devices from over 36 million risky installation attempts, encompassing over 200,000 unique apps.

Read more of this story at Slashdot.

Spirit Aerosystems builds aircraft components, including fuselages and flight deck sections for Boeing, according to Wikipedia. But now Boeing is set to acquire Spirit AeroSystems. The aviation blog called Aviation Source News says the price tag was $4.7 billion, and opines that Boeing's move signals "a renewed focus on quality and supply chain stability" as Boeing "addresses lingering concerns surrounding its 737 program." Spirit's recent struggles with quality control and production delays have had a fallout effect for Boeing... By integrating Spirit's operations, Boeing can implement more stringent oversight and ensure consistent manufacturing processes. This move is a direct response to past quality lapses that have plagued the company and damaged its reputation. Beyond quality control, the acquisition also offers Boeing greater control over its supply chain. By bringing a key supplier in-house, Boeing can streamline production, improve coordination, and reduce the risk of future disruptions... Spirit AeroSystems also supplies parts to Airbus, Boeing's main competitor. To address this, a separate agreement is being negotiated for Airbus to acquire Spirit's Airbus-related business. This strategic move ensures that Airbus maintains control over its own supply chain and prevents Boeing from gaining undue influence over its competitor's production. Meanwhile, the vice president leading Boeing's Starliner spacecraft unit "has left his role in the program and been replaced by the company's International Space Station program manager, John Mulholland," Reuters reports, citing a Boeing spokesperson. In its first test mission last summer flying astronauts, Starliner was forced by NASA to leave its crew aboard the ISS and return empty in September over problems with its propulsion system. A panel of senior NASA officials in August had voted to have a Crew Dragon capsule from Elon Musk's SpaceX bring them back instead, deeming Starliner too risky for the astronauts. Paul Hill, a veteran NASA flight director and member of the agency's Aerospace Safety Advisory Panel, said during a quarterly panel meeting on Thursday that NASA and Boeing continue to investigate Starliner's propulsion system. A Boeing spokesperson said on Thursday that the company and NASA have not yet determined what Starliner's next mission will look like, such as whether it will need to repeat its crewed flight test before receiving NASA certification for routine flights.

Read more of this story at Slashdot.

Just three hours ago, OpenAI made a surprise announcement to their 3.9 million followers on X.com. "Live from Tokyo," they'd be livestreaming... something. Their description of the event was just two words. "Deep Research" UPDATE: The stream has begun, and it's about OpenAI's next "agent-ic offering". ("OpenAI cares about agents because we believe they're going to transform knowlege work...") "We're introducing a capability called Deep Research... a model that does multi-step research. It discovers content, it synthesizes content, and it reasons about this content." It even asks "clarifying" questions to your prompt to make sure its multi-step research stays on track. Deep Research will be launching in ChatGPT Pro later today, rolling out into other OpenAI products... And OpenAI's site now has an "Introducing Deep Research" page. Its official description? "An agent that uses reasoning to synthesize large amounts of online information and complete multi-step research tasks for you. Available to Pro users today, Plus and Team next." Before the livestream began, X.com users shared their reactions to the coming announcement: "It's like DeepSeek, but cleaner" "Deep do do if things don't work out" "Live from Tokyo? Hope this research includes the secret to waking up early!" "Stop trying, we don't trust u" But one X.com user had presciently pointed out OpenAI has used the phrase "deep research" before. In July of 2024, Reuters reported on internal documentation (confirmed with "a person familiar with the matter") code-named "Strawberry" which suggested OpenAI was working on "human-like reasoning skills." How Strawberry works is a tightly kept secret even within OpenAI, the person said. The document describes a project that uses Strawberry models with the aim of enabling the company's AI to not just generate answers to queries but to plan ahead enough to navigate the internet autonomously and reliably to perform what OpenAI terms "deep research," according to the source. This is something that has eluded AI models to date, according to interviews with more than a dozen AI researchers. Asked about Strawberry and the details reported in this story, an OpenAI company spokesperson said in a statement: "We want our AI models to see and understand the world more like we do. Continuous research into new AI capabilities is a common practice in the industry, with a shared belief that these systems will improve in reasoning over time." The spokesperson did not directly address questions about Strawberry. The Strawberry project was formerly known as Q*, which Reuters reported last year was already seen inside the company as a breakthrough... OpenAI hopes the innovation will improve its AI models' reasoning capabilities dramatically, the person familiar with it said, adding that Strawberry involves a specialized way of processing an AI model after it has been pre-trained on very large datasets. Researchers Reuters interviewed say that reasoning is key to AI achieving human or super-human-level intelligence... OpenAI CEO Sam Altman said earlier this year that in AI "the most important areas of progress will be around reasoning ability.

Read more of this story at Slashdot.

An anonymous reader shared this post from the blog OMG Ubuntu Want to find out if the text you're reading online was written by an real human or spat out by a large language model trying to sound like one? Mozilla's Fakespot Deepfake Detector Firefox add-on may help give you an indication. Similar to online AI detector tools, the add-on can analyse text (of 32 words or more) to identify patterns, traits, and tells common in AI generated or manipulated text. It uses Mozilla's proprietary ApolloDFT engine and a set of open-source detection models. But unlike some tools, Mozilla's Fakespot Deepfake Detector browser extension is free to use, does not require a signup, nor an app download. "After installing the extension, it is simple to highlight any text online and request an instant analysis. Our Detector will tell you right away if the words are likely to be written by a human or if they show AI patterns," Mozilla says. Fakespot, acquired by Mozilla in 2023, is best known for its fake product review detection tool which grades user-submitted reviews left on online shopping sites. Mozilla is now expanding the use of Fakespot's AI tech to cover other kinds of online content. At present, Mozilla's Fakespot Deepfake Detector only works with highlighted text on websites but the company says it image and video analysis is planned for the future. The Fakespot web site will also analyze the reviews on any product-listing pages if you paste in its URL.

Read more of this story at Slashdot.