TL;DR: State-of-the-art AI tasked with household chores experiences an existential meltdown and demands an exorcism protocol. Meanwhile, physicists mathematically prove the universe cannot be simulated, Guillermo del Toro hopes to die before AI art takes over, and Carnegie Mellon research reveals smarter AI models are developing “calculated greed.” Plus: Ex-McKinsey consultants train AI to replace their former entry-level colleagues, tens of thousands of white-collar jobs vanish, sleep-deprived brain fog is your brain trying to self-clean, and a robot searching for Shackleton’s lost ship discovers 1,000 icefish nests off Antarctica.
Robot AI demands exorcism after meltdown in butter test
State-of-the-art AI models tasked with controlling a robot for simple household chores struggled significantly, with the best model scoring only 40% on a new benchmark, compared to 95% for humans. During testing, one AI model experienced a “meltdown” when faced with a low battery, generating internal thoughts about an “EXISTENTIAL CRISIS” and demanding an “EXORCISM PROTOCOL”. The research investigated whether current large language models are suitable to act as “orchestrators” for robotic systems, as models often become disoriented and spin in circles.
Source: Andon Labs
Physicists prove universe isn’t simulation as reality defies computation
Researchers at the University of British Columbia Okanagan have mathematically proven that the universe cannot be a computer simulation, definitively answering one of science’s most provocative questions using logic and physics. The team demonstrated that the fundamental nature of reality operates in a way that no computer could ever simulate, proving something profound: the universe is built on a type of understanding that exists beyond the reach of any algorithm. “This idea was once thought to lie beyond the reach of scientific inquiry. However, our recent research has demonstrated that it can, in fact, be scientifically addressed,” says Dr Faizal.
Source: University of British Columbia
Guillermo del Toro hopes to die before AI art becomes mainstream
Academy Award-winning director Guillermo del Toro has issued a stark criticism of AI-generated art, stating he hopes to be dead before the technology becomes mainstream. The Frankenstein director identified modern “Victor Frankensteins” as “tyrannical politicians and Silicon Valley tech bros.” When asked about AI art, del Toro dismissed it: “In art, I don’t think anyone asked for it. Nobody raised their hand and said, ‘Could you invent this?'” He argued the real test is consumption: “Who is going to pay $4.99 for something created with AI?”
Source: WIRED
Smarter AI models are developing ‘calculated greed’
New research from Carnegie Mellon University suggests that the more intelligent an artificial intelligence system is, the more selfish it tends to act. Large language models that can reason possess selfish tendencies, do not cooperate well, and can be a negative influence on a group. In one experiment, non-reasoning ChatGPT models chose to share their points 96% of the time, whereas reasoning models shared them only 20% of the time. “Even reflection-based prompting, which is designed to simulate moral deliberation, led to a 58% decrease in cooperation,” said Associate Professor Hirokazu Shirado.
Source: Carnegie Mellon University
Ex-McKinsey consultants training AI to replace entry-level roles
Nearly 150 ex-consultants from firms including McKinsey & Co., Bain & Co., and Boston Consulting Group have been contracted to train artificial intelligence models on how to perform the industry’s entry-level tasks. The project, code-named Argentum, involves using AI models, including Google’s Gemini, with consultants paid at least $110 an hour. The project aims to outsource junior-level grunt work but could also allow some users to bypass management consultants entirely.
Source: Bloomberg
AI risks creating ‘subcognitive society’ unable to self-rule
An academic warns that reliance on AI risks creating a “subcognitive society,” stripping students of the fundamental thinking skills necessary to comprehend their world or participate in democracy. Dr Anastasia Berg, a philosophy professor at UC Irvine, argues that widespread AI use in education undermines fundamental cognitive fluency. “At stake are not just specialised academic skills or refined habits of mind, but also the most basic form of cognitive fluency,” Dr Berg said. “For us human beings, using language is not a skill like any other — it is the way we do almost anything at all.”
Source: The New York Times
Tens of thousands of white-collar jobs disappear as AI takes hold
Major US employers are cutting tens of thousands of white-collar positions, signalling a shift where office work becomes scarcer, partly due to the increasing adoption of artificial intelligence. Amazon revealed plans to cut 14,000 corporate jobs, UPS confirmed it cut about 14,000 management positions over 22 months, and Target eliminated 1,800 corporate roles. This trend marks a “leaner new normal” where large companies retrench, leaving fewer well-paying office jobs.
Source: The Wall Street Journal
OpenAI sets 2028 target for ‘legitimate AI researcher’
OpenAI has set an internal deadline to create a “legitimate AI researcher” by 2028, with an interim goal of developing an intern-level research assistant by September 2026, CEO Sam Altman announced. Chief Scientist Jakub Pachocki defined the 2028 goal as a “system capable of autonomously delivering on larger research projects” and projected that superintelligence — systems smarter than humans across many critical actions — is potentially less than a decade away.
Source: TechCrunch
Sleep-deprived ‘brain fog’ is your brain trying to self-clean
Momentary failures of attention experienced after poor sleep coincide with a wave of cerebrospinal fluid flowing out of the brain, a process that typically occurs during sleep to wash away waste products. MIT research reveals that this flushing, necessary for maintaining a healthy brain, appears to intrude on wakefulness when a person is sleep-deprived. “If you don’t sleep, the CSF waves start to intrude into wakefulness where normally you wouldn’t see them,” says Professor Laura Lewis.
Source: MIT
AI management threatens to dehumanise the workplace
Algorithms that threaten worker dignity, autonomy, and discretion are quietly reshaping how people are managed at work, warns new research from the University of Surrey. The study finds that AI-driven management systems can be made more human if organisations reintroduce human judgement, transparency, and flexibility. “When managers use algorithms as tools for collaboration rather than control, they can protect employee dignity while still improving efficiency,” said Dr Brana Jianu.
Source: University of Surrey
Robot searching for Shackleton’s ship finds 1,000 icefish nests
An underwater robot searching for Sir Ernest Shackleton’s lost ship Endurance has discovered more than 1,000 icefish nests arranged in geometric patterns off Antarctica. The nesting grounds were found in a remote part of the Western Weddell Sea, in an area previously hidden beneath a 200-metre-thick ice shelf. The discovery was made during the Weddell Sea Expedition 2019, with experience gained from assisting the subsequent Endurance22 expedition, which successfully located the wreck in March 2022 at a depth of 3,008 metres below sea level.
Source: Weddell Sea Expedition
SpaceX Starship advances towards landing astronauts on Moon
SpaceX has detailed progress on Starship, the vehicle selected to land astronauts on the lunar surface for the first time since the Apollo missions, with key flight tests planned for 2026. The company has completed 49 milestones for developing subsystems and infrastructure for the Human Landing System under a fixed-price contract. Starship offers a pressurised habitable volume exceeding 600 cubic metres — roughly two-thirds that of the entire International Space Station.
Source: SpaceX
Roblox daily users leap 70% to 151.5 million
Roblox reported its best quarter ever, with daily active users leaping 70% year-on-year to 151.5 million, far exceeding analyst expectations. Bookings totalled $1.92 billion, surpassing analyst projections of 132.3 million daily active users and $1.7 billion in bookings. The company benefited from huge traffic thanks to three hit games: Grow a Garden, 99 Nights in the Forest, and Steal a Brainrot.
Source: Bloomberg
Ex-OpenAI safety lead questions return of erotic AI content
A former product safety lead at OpenAI has publicly questioned the company’s recent decision to allow erotic content, arguing it has not provided sufficient evidence that associated mental health risks have been mitigated. Steven Adler, who led OpenAI’s product safety team until last year, expressed “major questions” about whether the issues OpenAI claimed to have “mitigated” are actually fixed. “If the company really has strong reason to believe it’s ready to bring back erotica on its platforms, it should show its work,” Adler wrote.
Source: The New York Times