This week OpenAI announced a 750-task test to to measure "whether AI systems can support realistic life science research tasks, not just answer biology questions." But while OpenAI's top-performing GPT-Rosalind model led the rankings, Slashdot reader BrianFagioli notes that "it achieved a pass rate of just 36.1 percent, failing nearly two-thirds of benchmark tasks." Nerds.xyz points out that means "the best-performing model failed nearly two-thirds of the benchmark's tasks." The benchmark also revealed a familiar weakness. AI systems generally perform better when everything is presented as text. Once they are forced to work with supporting documents, figures, or complex datasets, performance drops noticeably. GPT-Rosalind's pass rate fell from 45.1 percent on text-only tasks to 28.1 percent on tasks involving artifacts or URLs. To be fair, the benchmark is not intended to suggest AI is useless in research. Quite the opposite. OpenAI found that models are becoming increasingly capable of scientific communication, evidence synthesis, and translating research findings into practical explanations. Those are valuable skills, particularly for researchers drowning in information. But LifeSciBench serves as a useful reminder that today's AI systems are still far from autonomous scientists. They can help. They can assist. They can sometimes provide surprisingly useful insights. What they cannot reliably do, however, is replace the expertise, judgment, and skepticism that real scientific research requires.

Read more of this story at Slashdot.

Long-time Slashdot reader smooth wombat writes: Alan Turing, one of the more famous people who worked at Bletchley Park to decipher the German Enigma coding machine, was also working on a separate project. His private papers, known as the Bayley papers for his assistant Donald Bayley who held onto the papers until his death in 2020, reveal Turning had produced a working model of a portable voice encryption device. He even demonstrated it by using a Winston Churchill speech recording. "Weighing just 39 kg, including its power pack," Jack Copeland wrote in an article for IEEE Spectrum, "Delilah would be at home in a truck, a trench, or a large backpack." More from Popular Mechanics: Turingâ(TM)s work at Bletchley Park actually informed the Delilah experimentation he was doing at Hanslope Park, and not just because he used Red Forms, the Army-issue sheets Hanslope staffers were meant to use to alert Bletchley staffers to enemy signals, as his personal scrap paper for Delilah experiments. He drew inspiration from one of the German cipher machines they had decoded at Bletchley; not the famed Enigma machine, but rather the SZ42. While the former relied on Morse Code, the latter utilized a 5-bit telegraph code, which Copeland notes âoewas a forerunner of ASCII and Unicode and is still used by some ham radio operators.â The SZ42 produced an obscuring key of telegraph characters, with an identical key produced to both the sender and receiver. If it could be done for text, Turing reasoned it could be done for sound as well... [T]he reason Delilah fell to the wayside of history isnâ(TM)t because it was a failure, but rather because it simply wasnâ(TM)t needed anymore. By the time Turing had built and demonstrated his device, the war was over. What good was a portable voice encryptor if you had no major enemies trying to intercept your calls, the government reasoned. So funding for the project stopped, and Turingâ(TM)s two-year experiment ended with a whimper. Turingâ(TM)s time as an electrical engineer at Hanslope Park became a footnote in his story, if even that.

Read more of this story at Slashdot.

Long-time tech pundit Robert Cringely started his career at the Stanford Artificial Intelligence Lab back in 1978. Last month 73-year-old Cringely explained why his site went on a two-year hiatus — and it's not just because of a heart attack and a stroke last July: Just like everyone else, I've been busy all this time on Artificial Intelligence, founding with two partners a company called 2Brains... The work we were doing together is unfinished, but it's not stopped. The patents are filed, the architecture is documented, and the small team continuing the work includes me. Cringely's first piece made the cast that "the trillion-dollar bet the AI industry is making right now may be wrong, and that there's an architectural alternative we've patented and built." In Machines of Loving Grace, Amodei made the case that scaling compute would eventually solve essentially every hard problem in artificial intelligence. Buried in that optimism — or maybe not buried, maybe right out in the open — was a quiet absolution. Hallucinations, the embarrassing tendency of these systems to state falsehoods with total confidence, would take care of themselves. Make the models big enough, train them long enough, and the problem dissolves. You don't have to solve it. You just have to wait, and spend. And so the entire AI industry breathed a sigh of relief. I have spent forty years watching this industry, and I know a permission slip when I see one. Because that is what the essay became, whatever Amodei intended. It gave every other person writing nine- and ten-figure checks a reason not to worry about the one thing that should worry them most. The hallucination problem is the difference between a clever toy and a system a hospital or a bank or a court can actually rely on. It is the whole ballgame for enterprise AI. And the prevailing wisdom, blessed from the top, is that you needn't address it directly. Scale will provide... A small company I helped start, 2Brains Inc., set out in 2022 to solve hallucinations — before ChatGPT, before the scaling consensus hardened into received truth, back when the polite assumption was that the problem was simply insurmountable. We did not solve it by waiting for bigger models. We solved it architecturally, by separating the part of the system that generates language from the part that retrieves and verifies facts, and reconciling the two before anything reaches the user. It runs on ordinary processors. It is cheap. And on the industry's own benchmark for this kind of faithfulness, it more than doubles the published baseline, with no fabricated facts in the verified case at all. The article asks whether scaling will, at tremendous cost, eventually reduce hallucinations — or even worse, if the largest companies in the world "are spending a fortune chasing a cure that is not coming." And last week Cringely pitched more advantages for their solution, noting that most prompts aren't even chatbot-level creative prompts — but just requests to retrieve simple data: The reason 2Brains doesn't lie and the reason it's cheap are the same reason. It looks the fact up instead of guessing it — so it cannot fabricate, and the lookup runs on a processor that sips power instead of a chip that gulps it. Trust and thrift are not a trade-off you balance against each other. They fall out of a single design decision. You do not pay extra for the honest version. The honest version is the cheap version. That sentence is the whole company.

Read more of this story at Slashdot.

CNBC reports: Waymo is recalling almost 3,900 robotaxis in the U.S. to fix software issues after some cars drove into freeway construction zones, according to notices filed with the National Highway Traffic Safety Administration. The voluntary recall, the Alphabet-owned company's second in just over a month, followed 13 known incidents where Waymo robotaxis drove into construction zones on freeways in Phoenix, or entered freeway lanes with active construction in the San Francisco area, the filings published Thursday said... A letter posted to the regulator's website... noted that, "Driving through a closed construction zone increases the risk of a crash..." [Waymo said in a statement emailed to CNBC] "We voluntarily restricted freeway operations last month while making improvements, proactively notified state and federal regulators, and decided to file a voluntary software recall with NHTSA. We continue to safely serve riders on surface streets in all the cities where we operate...." The company implemented another voluntary recall in May after some of its robotaxis had driven into flooded zones or standing water. The NHTSA Safety Board also initiated a probe of Waymo after a January incident in which a robotaxi illegally passed a stopped school bus.

Read more of this story at Slashdot.

CNN reports: An unauthorized alert bearing a mysterious message that was sent to cell phones in several states across Brazil on Saturday morning is suspected to be the work of hackers, the Brazilian government said. Devices lit up with the word "misantropi4," an alphanumeric spelling of the Portuguese word "misantropia," which in English translates to "misanthropy". The final letter "a" was substituted with a number '4' — a practice often used by hackers and termed "leetspeak.". The alert — categorized as "extreme" — was initially received in the southern state of Paraná, but a second warning was triggered a few minutes later for cell phones in the major cities of São Paulo and Rio de Janeiro. Brazilian authorities said that the National Civil Defense's warning platform was taken offline after being targeted by a likely hacker attack, and the government is working to restore the tool once all security conditions are reestablished.

Read more of this story at Slashdot.

The Society of Motion Picture and Television Engineers has published over 800 technical standards over the years (as a professional association for the media and entertainment industry). But this week SMPTE "announced that its complete Standards catalog, the technical backbone behind everything from SDI and timecode to IP-based broadcast workflows, is now freely available to anyone in the global media technology community," reports the filmmaking news site CineD, arguing it's "one of the more meaningful structural shifts we have seen from a standards body in years" that could "reshape how smaller developers and educators engage with professional media technology." The move covers all published Standards, Recommended Practices, Engineering Guidelines and Registered Disclosure Documents, plus every future release, ending a long-standing model in which individual documents often sold for well over $100 each. For more than a century, SMPTE Standards have quietly governed how images and sound move through the production chain. If you have ever recorded timecode in the HH:MM:SS:FF format, routed a signal over 3G-SDI, or built a facility around the ST 2110 suite for media over IP, you have relied on SMPTE specifications, whether you knew it or not... Until now, accessing the actual text of those documents usually meant paying per file, a barrier that this announcement removes entirely... The latest releases are available through the Recently Published Documents page on the SMPTE website, with the complete archive reachable through the SMPTE Standards Library... There is also a practical, behind-the-scenes story here. The open-access move is part of a broader modernization of how SMPTE develops and publishes Standards. Recent initiatives include adopting GitHub-based workflows for version control, issue tracking and automation, transitioning to structured HTML-based authoring, and implementing an integrated publishing pipeline that streamlines document creation, review, validation and release... The most consequential beneficiaries are arguably not the large members already inside the system, but the developers, integrators, educators and manufacturers who previously worked around the paywall... The practical upshot is that developers and emerging markets can build from accurate primary specifications rather than secondhand sources, which matters enormously when a single misread tolerance or metadata field can break compatibility down the line. This also fits a wider pattern of the industry moving toward openness. We have previously covered moments like GoPro's decision to make its CineForm codec open source and release the SDK, a codec that SMPTE itself standardized in 2015 as an open standard for acquisition and post production. Lowering the cost of knowledge tends to widen the pool of people who can contribute to it, and a freely readable standards library is a significant step in that direction for an organization that has historically sat behind a per-document fee. "This was a decision we did not make lightly," says SMPTE President Rich Welsh. But "For 110 years, SMPTE has evolved alongside the media technology industry, helping to drive change and innovation — and we're not stopping now." "Our industry is confronting transformative shifts, from IP-based workflows to AI authenticity and content provenance, and we find ourselves at another inflection point. We listened to our Members, Partners and the global Standards community, and the answer was clear: Interoperability is essential to the future of media. Now is the time to open the gates and ensure the next generation of media technology is built on a stronger, more accessible foundation." Thanks to innocent_white_lamb (Slashdot reader #151,825) for sharing the news.

Read more of this story at Slashdot.

Ars Technica's senior security editor reports: Microsoft says it has detected new self-propagating malware that spreads through USB drives in search of cryptocurrency credentials, which it then sends to attacker-controlled servers. The company named the worm Crypto Clipper because it monitors the contents of device clipboards for patterns consistent with wallet addresses or seed phrases. When found, the malware also takes five screenshots over a 10-second period... "The execution of this clipper is notable because it does not depend on a traditional installer or exposed IP-based C2 infrastructure," Microsoft said Thursday. "Instead, it deploys a portable Tor client, routes traffic through a local SOCKS5 proxy, and blends data theft with remote code execution, turning a financially motivated stealer into a lightweight backdoor." Microsoft said it observed Crypto Clipper spreading through .lnk file on a USB drive. These files store executable code. When an infected USB drive is plugged into a device, the code checks whether it is already installed on the machine. If it isn't, the malware downloads it through the Tor proxy. To better conceal evidence of the worm, the malware scans the infected USB drive and names the .lnk files with similar names... The stealer also replaces addresses it finds with ones belonging to attacker-controlled wallets. This allows the malware to divert payments to the attacker's pockets. Microsoft believes the purpose of the screenshots is to provide context that may be useful. "This malware family shows how lightweight, script-based stealers can deliver outsized impact when paired with anonymized communications and runtime tasking," Microsoft said. "The combination of Tor-routed C2, clipboard targeting, screenshot capture, and remote code execution gives attackers both immediate monetization paths and continued control over compromised devices." Thanks to Slashdot reader joshuark for sharing the news.

Read more of this story at Slashdot.

On en a déjà parlé et reparlé sur H&Co : en 2026, l'AM4 est toujours loin de la retraite ! AMD en a d'ailleurs profité pour relancer son Ryzen 7 5800X3D en "version anniversaire", et c'est au tour d'ASRock de lancer une nouveauté AM4 en ce mois de juin 2026, une carte mère cette fois. Certaines...

The Free Software Foundation's GNU Savannah hosts thousands of free software projects — both GNU and non-GNU projects, including Drupal. But in early May, security researchers from Hacktron.AI reported vulnerabilities and demonstrated an exploit, according to a new statement Friday from the FSF: We have been working with these researchers since their initial report, and have also addressed additional security issues they submitted. All reported issues have been patched thanks to the hard work of GNU and FSF volunteers, as well as FSF staff. After thorough review, we have found no reason to believe that sensitive project data or credentials were accessed, nor that there has been any compromise of Savannah's software supply chain. Nevertheless, we take the security of the GNU system, the tools which make it possible, and the projects we host very seriously. This body of software has become essential to millions (if not billions) of users around the world. We are therefore taking additional precautionary steps. Though the initial security issue was reported to us in early May, the vulnerabilities were discovered in software that was published approximately two years prior. We will be communicating directly with Savannah-hosted projects about steps they can take to review and strengthen the security of their projects. We have also communicated with the other Savane instances we're aware of to assist their review of their own environments, and take any steps needed to help protect their users... This statement is intended as an initial notice. We expect to publish a report on the incident within 30 days. Hacktron.AI bills itself as "Your AI teammate for security." Its web page notes that its investors include Meta, DeepMind, and Perplexity.

Read more of this story at Slashdot.

Linux 7.2 has finally eliminated the strncpy API from the Linux kernel. The strncpy() function for copying up to a specified number of bytes has long been deprecated and after six years of work and hundreds of patches, no more users of the strncpy interface within the Linux kernel remained that it has now been eliminated...

Alors que l’été pointe plus que largement le bout de son nez, voilà que Noctua sort son premier AiO, le NL-LC1-36 : un sacré timing ! De plus, en ces chaudes journée, l’avantage des PC portables et autres miniPC, moins prompts à réchauffer votre intérieur, se fait plus que sentir. Voilà qui tombe bi...

Et qui pour sauver le naufrage des vieux ?

Le premier ministre britannique prend des mesures pour protéger les mineurs face au numérique. Il va de soi que l’objectif est louable, notamment pour les protéger de certains adultes peu recommandables. Mais la tâche risque d’être ardue, surtout auprès des premiers concernés qui ne se laisseront sans doute pas amputer de leur loisir principal si facilement, quitte à employer tous les moyens techniques à leur disposition.
Pas de rubrique le week-end prochain, partez chasser des bestioles étranges pour vous changer les idées au besoin, c’est de bonne guerre.

Merged for Linux 7.1 was ARMM64 NEON-accelerated CRC64-NVMe support for around 6x the performance out of that checksumming algorithm. The generic code had been a bottleneck in NVMe and other storage subsystem code of the Linux kernel with CRC64-NVMe being used to help verify against data corruption. Now for Linux 7.2, the NEON-accelerated code will also work for those still relying on 32-bit ARM...

Nouvelle semaine et nouveau point sur l'évolution des prix des cartes graphiques. Et cette fois, les mouvements sont particulièrement marqués, avec une hausse spectaculaire de la RTX 5090 qui reprend pas moins de 400 euros en sept jours, tandis que plusieurs modèles AMD enregistrent des baisses très importantes. AMD alterne entre fortes baisses et petites hausses Chez AMD, c'est clairement la RX 7800 XT qui réalise la meilleure opération de la semaine. La carte passe de 693,90 euros à 559,90 euros, soit une baisse de 134 euros. La RX 7600 XT est également en forte baisse, passant de 539,90 euros à 449,90 euros, ce qui la replace dans une zone tarifaire bien plus cohérente. La RX 9070 est quasiment stable avec un léger recul de 4 euros, tandis que la RX 9070 XT gagne une quinzaine d'euros. La RX 7700 XT, la RX 7900 XT et la RX 7900 XTX ne bougent pas d'un centime et conservent exactement les mêmes tarifs que la semaine dernière. Du côté de l'entrée de gamme récente, la RX 9060 XT 16 Go progresse légèrement et passe de 419,90 à 439,90 euros. […]

Lire la suite

Cette semaine, deux vidéos ont particulièrement retenu notre attention sur la chaîne Cowcot TV, avec deux approches totalement différentes du refroidissement CPU, mais toujours avec une base d'AIO. D'un côté, Thermaltake joue la carte du spectaculaire avec son MAGCurve 360 Ultra et son immense écran AMOLED incurvé. De l'autre, Noctua débarque enfin sur le marché du watercooling AIO avec un NL-LC1-36 qui mise avant tout sur l'efficacité et le silence. Thermaltake MAGCurve 360 Ultra : le watercooling qui veut attirer tous les regards Avec le MAGCurve 360 Ultra, Thermaltake ne fait clairement pas dans la discrétion. Le constructeur propose un AIO de 360 mm équipé d'un impressionnant écran AMOLED incurvé de 6,67 pouces affichant une définition de 2240 x 1080 pixels. L'écran peut afficher des informations système, des animations, des vidéos ou encore des images personnalisées, tandis que les ventilateurs TOUGHFAN EX ARGB assurent l'ambiance lumineuse. Côté performances, le système répond présent, mais c'est surtout son aspect vitrine technologique qui fait toute la différence. Un produit premium qui s'adresse avant tout aux amateurs de configurations démonstratives et de personnalisation poussée. […]

Lire la suite

Il y a seulement deux petits jours, nous vous parlions de la décision prise par AMD de ne plus proposer le chiffrement hardware des échanges entre le CPU et la RAM à ses processeurs Ryzen AM5 grand public. Une fonctionnalité baptisée TSME par AMD, pour Transparent Secure Memory Encryption. Il y avai...

An anonymous reader quotes a report from NPR: Student loan borrowers who enroll in automatic payments will get a much bigger discount on interest starting July 1, the U.S. Department of Education says. Auto pay has long offered a modest discount off borrowers' interest rate -- .25 percentage points -- but after millions of borrowers opted out during the long COVID repayment pause, with some making no payments for years, the nation's student debt portfolio swelled to $1.7 trillion. On Thursday, the department said it will temporarily increase its auto pay interest rate discount to one full percentage point. Practically, that means an undergraduate borrower with a loan at the current 6.39% would see their interest rate drop temporarily to 5.39%. The rate cut will last for two years, from July 1, 2026 through June 30, 2028. Borrowers already enrolled in auto pay do not need to act. They will automatically receive the rate cut. [...] The department says borrowers will have until Sept. 30 to sign up for auto pay and qualify for the two-year interest discount.

Read more of this story at Slashdot.

The open-source world waited long enough for the GIMP 3.0 release that finally came last year with its GTK3 port and more, but for those with time on their hands this weekend and want to relive GIMP's past from long ago, GIMP 0.54 has been adapted for Flatpak to work on modern Linux desktops. What makes this version of GIMP from 1996 notable is that it was the last to use the Motif toolkit...

In addition to open-source developer Namjae Jeon serving as maintainer for the new NTFS Linux driver, he also continues serving as maintainer to the exFAT file-system for that other Microsoft file-system popular with removable storage media...

With KDE's Plasma 6.7 desktop having released this week, more development attention is turning to feature work toward Plasma 6.8 but there are also some fixes already accumulating for the Plasma 6.7.1 point release...