Linux fréttir

Don’t laugh, a French nuclear power plant just shut down for a while after invertebrates overwhelmed its intakes

Proponents of increased use of nuclear energy to power datacenters have a new foe: Jellyfish.…

An anonymous reader quotes a report from Ars Technica: In recent months, the AI industry has started moving toward so-called simulated reasoning models that use a "chain of thought" process to work through tricky problems in multiple logical steps. At the same time, recent research has cast doubt on whether those models have even a basic understanding of general logical concepts or an accurate grasp of their own "thought process." Similar research shows that these "reasoning" models can often produce incoherent, logically unsound answers when questions include irrelevant clauses or deviate even slightly from common templates found in their training data. In a recent pre-print paper, researchers from the University of Arizona summarize this existing work as "suggest[ing] that LLMs are not principled reasoners but rather sophisticated simulators of reasoning-like text." To pull on that thread, the researchers created a carefully controlled LLM environment in an attempt to measure just how well chain-of-thought reasoning works when presented with "out of domain" logical problems that don't match the specific logical patterns found in their training data. The results suggest that the seemingly large performance leaps made by chain-of-thought models are "largely a brittle mirage" that "become[s] fragile and prone to failure even under moderate distribution shifts," the researchers write. "Rather than demonstrating a true understanding of text, CoT reasoning under task transformations appears to reflect a replication of patterns learned during training." [...] Rather than showing the capability for generalized logical inference, these chain-of-thought models are "a sophisticated form of structured pattern matching" that "degrades significantly" when pushed even slightly outside of its training distribution, the researchers write. Further, the ability of these models to generate "fluent nonsense" creates "a false aura of dependability" that does not stand up to a careful audit. As such, the researchers warn heavily against "equating [chain-of-thought]-style output with human thinking" especially in "high-stakes domains like medicine, finance, or legal analysis." Current tests and benchmarks should prioritize tasks that fall outside of any training set to probe for these kinds of errors, while future models will need to move beyond "surface-level pattern recognition to exhibit deeper inferential competence," they write.

Read more of this story at Slashdot.

Outages, degraded service, and login troubles hit 10 regions and 27 services

IBM Cloud experienced a Severity One outage on Monday that left customers unable to access resources.…

AmiMoJo shares a report from the BBC: A French nuclear plant temporarily shut down on Monday due to a "massive and unpredictable presence of jellyfish" in its filters, its operator said. The swarm clogged up the cooling system and caused four units at the Gravelines nuclear power plant to automatically switch off, energy group EDF said. The plant is cooled from a canal connected to the North Sea -- where several species of jellyfish are native and can be seen around the coast when the waters are warm. According to nuclear engineer Ronan Tanguy, the marine animals managed to slip through systems designed to keep them out because of their "gelatinous" bodies. "They were able to evade the first set of filters then get caught in the secondary drum system," he told the BBC. Mr Tanguy, who works at the WNA, said this will have created a blockage which reduced the amount of water being drawn in, prompting the units to shut down automatically as a precaution. He stressed that the incident was a "non-nuclear event" and more a "nuisance" for the on-site team to clean up. For local people, there would be no impact on their safety or how much energy they could access: "They wouldn't perceive it as any different to any other shut-down of the system for maintenance."

Read more of this story at Slashdot.

Legal scholar Victoria Haneman argues that U.S. law should grant estates a time-limited right to delete a deceased person's data so they can't be recreated by AI without their consent. "Digital resurrection by or through AI requires the personal data of the deceased, and the amount of data that we are storing online is increasing exponentially with each passing year," writes Haneman in an article published earlier this year in the Boston College Law Review. "It has been said that data is the new uranium, extraordinarily valuable and potentially dangerous. A right to delete will provide the decedent with a time-limited right for deletion of personal data." The Register reports: A living person may have some say on the matter through the control of personal digital documents and correspondence. But a dead person can't object, and US law doesn't offer the dead much data protection in terms of privacy law, property law, intellectual property law, or criminal law. The Revised Uniform Fiduciary Access to Digital Assets Act (RUFADAA), a law developed to help fiduciaries deal with digital files of the dead or incapacitated, can come into play. But Haneman points out that most people die intestate (without a will), leaving matters up to tech platforms. Facebook's response to dead users is to allow anyone to request the memorialization of an account, which keeps posts online. As for RUFADAA, it does little to address digital resurrection, says Haneman. The right to publicity, which provides a private right of action against unauthorized commercial use of a person's name, image, or likeness, covers the dead in about 25 states, according to Haneman. But the monetization of publicity rights has proven to be problematic. Haneman says that there are some states where it's theoretically possible to be prosecuted for libeling or defaming the deceased, such as Idaho, Nevada, and Oklahoma, but adds that such prosecutions have declined because they tread upon the constitutional right to free expression. [...] A recent California law, the Delete Act, which took effect last year, is the first to offer a way for the living to demand the deletion of personal data from data brokers in one step. But according to Haneman, it's unclear whether the text of the law will be extended to cover the dead -- a possibility think tank Aspen Tech Policy Hub supports [PDF]. Haneman argues that a data deletion law for the dead would be grounded in laws governing human remains, where corpses receive protection against abuse despite being neither a person nor property. "The personal representative of the decedent has the right to destroy all physical letters and photographs saved by the decedent; merely storing personal information in the cloud should not grant societal archival rights," she argues. "A limited right of deletion within a twelve-month window balances the interests of society against the rights of the deceased."

Read more of this story at Slashdot.

Just days after demanding Intel CEO Lip-Bu Tan resign over his past ties to China, President Trump reversed course, calling Tan a "success" following a White House meeting. "I met with Mr. Lip-Bu Tan, of Intel, along with Secretary of Commerce, Howard Lutnick, and Secretary of the Treasury, Scott Bessent," Trump wrote in a post on Truth Social. "The meeting was a very interesting one. His success and rise is an amazing story. Mr. Tan and my Cabinet members are going to spend time together, and bring suggestions to me during the next week. Thank you for your attention to this matter!" CNBC reports: Tan has been an Intel director since 2022, and in March he replaced Pat Gelsinger as CEO. Last week Sen. Tom Cotton, R-Ark., questioned Tan's ties to China. Cotton brought up a past criminal case involving Cadence Design, where Tan had been CEO, and asked whether Intel required Tan to divest from positions in chipmakers linked to the Chinese Communist Party, the People's Liberation Army and any other concerning entities in China. Trump's latest message marks a stark change in tone from last week. In a Truth Social post on Thursday, the president wrote that Tan "is highly CONFLICTED and must resign, immediately. There is no other solution to this problem." Intel said in a comment later that day that the company, directors and Tan are "deeply committed to advancing U.S. national and economic security interests."

Read more of this story at Slashdot.

70W TDP means the new RTX Pro 4000 SFF and RTX Pro 2000 won't blow power budgets

Nvidia’s latest Blackwell GPUs are a pair of itty-bitty workstation cards that aim to deliver the highest performance possible for professional visualization and local AI workloads within a 70-watt energy diet.…