"Every link leads to an entry that does not exist yet," explains the GitHub page for a Wikipedia-like site called Halupedia. "Until you click it, at which point an LLM pretends it has always existed and writes it for you, in the deadpan register of a 19th-century scholarly press..." Every article is invented on demand. The footnotes are also lies... The hardest problem with an infinite, on-demand encyclopedia is internal contradiction... When the LLM writes an article, it is required to add a context="..." attribute on every <a> it inserts, summarising the future article it is linking to (e.g. context="19th-century clerk who formalized footnote drift, Pellbrick's mentor")... When that target article is later requested for the first time, the worker loads the accumulated hints and injects them into the system prompt as "PRIOR REFERENCES — these are CANON". The LLM is instructed that the encyclopedia is hallucinated and absurd, but it must not contradict itself. Fast Company reports that Halupedia was created by software developer Bartlomiej Strama, who confessed in a Reddit comment that the site came about after a drunk night with a friend. In the week since launch, he says Halupedia has amassed more than 150,000 users." Beyond indulging in silly alternate histories, what's the point of using Halupedia? Strama hinted at one larger purpose in a reply to a donor on his Buy Me a Coffee page: "Your contribution towards polluting LLM training data will surely benefit society!" he wrote. The site is licensed as free software under the GPL-3.0 license. Thanks to long-time Slashdot reader schwit1 for sharing the news.

Read more of this story at Slashdot.

Bloomberg reports that Apple's two-year-old partnership with OpenAI "has become strained, according to people familiar with the matter." Bloomberg describes OpenAI as "failing to see the expected benefits from the deal and now preparing possible legal action." OpenAI lawyers are actively working with an outside legal firm on a range of options that could be formally executed in the near future, said the people, who asked not to be identified because the deliberations are private. That could include sending the iPhone maker a notice alleging breach of contract without necessarily filing a full lawsuit at the outset, according to the people... OpenAI believed that the companies' partnership, which wove ChatGPT into Apple software, would coax more users into subscribing to the chatbot. It also expected deeper integration across more Apple apps and prime placement within the Siri assistant. Instead, Apple's use of OpenAI technology across its operating systems remains limited, and features can be hard to find... Apple has had its own concerns about OpenAI, including whether the company does enough to protect user privacy. And a recent push [by OpenAI] to make devices — an effort overseen by former Apple executives — has rankled the iPhone maker. Any legal move by OpenAI likely wouldn't come until after the conclusion of the Musk trial, according to the people. No final decisions have been made, and OpenAI still hopes to resolve its issues with Apple outside of court. The article points out that OpenAI "initially believed the deal could generate billions of dollars per year in subscriptions — something that hasn't come close to happening." An OpenAI executive argues to Bloomberg that from a product perspective Apple hasn't done everything they could, "and worse, they haven't even made an honest effort."

Read more of this story at Slashdot.

"The vulnerability is simple in practice," writes Tom's Hardware: "run a command as a standard user and gain root (administrator) access to the machine." And it was Mythos Preview that helped the security researchers at Palo Alto-based Calif bypass a five-year Apple security effort in just five days. The blog 9to5Mac reports: Last year, Apple introduced Memory Integrity Enforcement (MIE), a hardware-assisted memory safety system designed to make memory corruption exploits much harder to execute... [The researchers note it's built into Apple all models of the iPhone 17 and iPhone Air, and some MacBooks] They explain they have a 55-page technical report on the hack, but they won't release it until Apple ships a fix for the exploit. But they do note in broad terms that Anthropic's Mythos Preview model helped them identify the bugs and assisted them throughout the entire collaborative exploit development process. "Mythos Preview is powerful: once it has learned how to attack a class of problems, it generalizes to nearly any problem in that class. Mythos discovered the bugs quickly because they belong to known bug classes. But MIE is a new best-in-class mitigation, so autonomously bypassing it can be tricky. This is where human expertise comes in. Part of our motivation was to test what's possible when the best models are paired with experts. Landing a kernel memory corruption exploit against the best protections in a week is noteworthy, and says something strong about this pairing...." [I]n a time when even small teams, with the help of AI, can make discoveries such as this one, "we're about to learn how the best mitigation technology on Earth holds up during the first AI bugmageddon."

Read more of this story at Slashdot.

Aucun historique de conversation stocké sur des serveurs : Mark Zuckerberg annonce un mode 100 % privé pour Meta AI sur WhatsApp. Une opération séduction pour rassurer des internautes légitimement méfiants à l'égard de l'IA et des chatbots, mais aussi envers le passif de l'entreprise américaine.

The Linux 7.1 kernel has added new documentation clarifying what qualifies as a security bug and how AI-assisted vulnerability reports should be handled. Phoronix reports: Stemming from the recent influx of security bugs to the Linux kernel as well as an uptick in bug and security reports from discoveries made in full or in part with AI, additional documentation was warranted. Longtime Linux developer Willy Tarreau took to authoring the additional documentation around kernel bugs. To summarize (since the documentation is a bit too lengthy for a Slashdot story), the AI-assisted vulnerability reports should "be treated as public" because such findings "systematically surface simultaneously across multiple researchers, often on the same day." It adds that reporters should avoid posting a reproducer openly, instead "just mention that one is available" and provide it privately if maintainers request it. The guidance also tells AI-assisted reporters to keep submissions concise and plain-text, focus on verifiable impact rather than speculative consequences, include a thoroughly tested reproducer, and, where possible, propose and test a fix. As for what qualifies as a security bug, the documentation says the private security list is for "urgent bugs that grant an attacker a capability they are not supposed to have on a correctly configured production system" and are easy to exploit, creating an imminent threat to many users. Reporters are told to consider whether the issue "actually crosses a trust boundary," since many bugs submitted privately are really ordinary defects that belong in the normal public reporting process. All the new documentation can be read via this commit.

Read more of this story at Slashdot.

Aucun historique de conversation stocké sur des serveurs : Mark Zuckerberg annonce un mode 100 % privé pour Meta AI sur WhatsApp. Une opération séduction pour rassurer des internautes légitimement méfiants à l'égard de l'IA et des chatbots, mais aussi envers le passif de l'entreprise américaine.