Author: lethuha

📌 Summary: The rate of workers using AI on the job in the US is sharply declining. By October 2025, only 11% of workers at large US companies used AI for “producing goods and services.” In medium-sized businesses, the percentage of those “not using AI” increased from 74.1% in March to 81.4%. The reason is attributed to “AI fatigue”: employees feel overwhelmed, while the resulting productivity gains are not commensurate.

📌 Summary: A child’s brain develops 90% before age 5, so early exposure to “AI slop” – low-quality, repetitive, and meaningless generative content – can affect language, cognitive, and emotional development. YouTubers are using generative AI tools like ChatGPT and automated video generators to mass-produce “educational” videos for children aged 1–3, earning hundreds of USD daily. With over 60% of children under 2 watching YouTube, “AI slop” is becoming a potential risk to brain development. While the platform commits to moderation, the reality shows that the algorithm still prioritizes profit and views over educational quality.

📌 Summary: The data center construction industry in Japan is facing severe pressure due to labor shortages, soaring costs, and outdated building processes, causing delays in the government’s AI infrastructure development plan. Japan currently faces two main issues: (1) a shortage of highly skilled workers, and (2) slow adoption of Building Information Modeling (BIM), which helps shorten approval and construction times. In Singapore, BIM has been mandatory since 2015, helping complete a 50 MW center in 2 years; in Japan, the same scale takes almost 4 years.

📌 Summary: The telecommunications operator Telenor Pakistan, in partnership with Data Vault, inaugurated the country’s first domestically hosted AI data center, integrating a sovereign AI cloud with 3,000 GPUs and offering GPU-as-a-Service. This marks a breakthrough enabling Pakistan to achieve AI infrastructure self-sufficiency and accelerate applications in finance, healthcare, public security, and agriculture. The “sovereign GPU cloud” infrastructure helps maintain all data within Pakistani territory.

📌 Summary: With Mistral 3, Europe launches a comprehensive open-source AI ecosystem that can run on everything from phones, drones, and robots to enterprise systems. The 10 diverse models – ranging from 3 billion to 675 billion parameters – allow AI to run “everywhere,” helping businesses reduce costs, increase autonomy, and pave the way for an era of distributed intelligence, where AI is not just big – but ubiquitous.

📌 Summary: The South Korean government terminated the “AI Digital Textbook Promotion Plan” just four months after implementation, following the chaos the program caused in schools. The government invested $850 million, and publishers contributed an additional $567 million to produce and deploy the books across 4,095 schools nationwide. However, the AI textbooks were plagued with serious flaws upon teaching. The program ended in failure: over 50% of schools withdrew, thousands of technical errors occurred, and teachers and students were exhausted. The incident highlights the consequences of hastily applying AI in education.

📌 Summary: The Australian Department of Home Affairs is establishing a new security platform to allow government agencies to use generative AI with sensitive data, classified from “OFFICIAL” up to “PROTECTED” level. Currently, only 18 AI vendors, including OpenAI (ChatGPT), Anthropic (Claude), and Google (Gemini), are permitted to access “OFFICIAL” data. The Department of Finance recently announced a policy requiring all public servants to have access to safe generative AI directly on their laptops. All government agencies must appoint a Chief AI Officer before July 2026 to oversee security policy compliance and practical implementation.

Contrary to fears that AI will replace humans, leading law firms like Allens and King & Wood Mallesons (KWM) are recruiting more lawyers to verify and supervise the results generated by AI. According to a survey by The Australian Financial Review, 70% of the 53 law firms questioned reported an increase in non-partner lawyers by more than 5%, reaching nearly 18,800 people, while the number of partners grew by 3% to 4,600. At Allens, the number of non-partner lawyers increased by 11% to a record high of 1,078, despite a slight decrease in partners to 157. The firm uses over 6 AI tools, including Arlie (an internal version of…

📌 Summary: Contrary to fears that AI will replace humans, leading law firms in Australia are recruiting more lawyers to verify and supervise the results generated by AI. The race by law firms to build proprietary “AI tech stacks” is creating a new wave of demand for lawyers, as courts require meticulous verification of AI outputs. Instead of cutting staff, AI is redefining the legal profession, transforming lawyers into “AI trainers”—both performing legal checks and leveraging technology to make more precise and creative decisions.

📌 Summary ICLR 2026—the world’s premier machine learning conference—is facing a shock as 21% of its peer reviews, equivalent to nearly 16,000 reviews, were found to be written entirely by generative AI. The received reviews contained incorrect data, bizarre phrasing, and missed the central focus of the research, which are characteristic features of feedback generated by generative AI.