Breaking News: Ohio State University to Mandate AI Training for All Students, Aiming for “Proficiency” in Artificial Intelligence
The Ohio State University has announced the launch of the “AI Fluency Initiative,” which requires all students, regardless of their major, to undergo training in artificial intelligence (AI) and achieve a level of "proficiency."
This program will begin this fall for incoming freshmen and is designed to ensure that graduates not only use AI tools effectively in their respective fields but also possess the ability to understand, question, and creatively apply AI. Additionally, the initiative aims to cultivate a deep understanding of AI ethics and secure usage.
Study Warns: Advanced AI “Completely Collapses” in Accuracy When Facing Complex Problems
A recent study has found that even so-called "advanced" AI models experience a complete collapse in accuracy when confronted with truly complex problems. This indicates that current AI systems still face significant limitations in solving problems that require deep logical reasoning and multi-step thinking.
The study noted that as the complexity of problems increases, AI models' performance does not decline linearly but rather experiences a sudden and sharp "collapse." This suggests that once a certain complexity threshold is reached, these models struggle to generate any correct solutions. For example, in problems like the "Tower of Hanoi" which require multiple steps of planning and precise reasoning, AI accuracy drops dramatically as difficulty increases.
TSMC's AI Chip Demand "Exceeds Supply," Pushing for Record Growth in 2025
Despite facing U.S. tariffs and global economic fluctuations, Taiwan Semiconductor Manufacturing Company (TSMC) expects 2025 to be a record-breaking year, driven largely by the sustained strong demand for artificial intelligence (AI) chips, which has even outpaced supply capabilities.
TSMC’s sales data for April highlighted this robust demand, with reported revenue of NT$349.6 billion (US$11.6 billion), marking a 48.1% increase year-over-year and a 22.2% growth compared to March.
OpenAI Expands Rapidly in South Korea, Seoul to Become Its Third Headquarters in Asia
AI giant OpenAI is accelerating its strategic expansion in South Korea, which has become its second-largest paid market globally. The user base of ChatGPT in the country has grown rapidly, with high engagement.
OpenAI has chosen Seoul to establish its third headquarters in Asia, recognizing the country’s unique AI ecosystem and enormous market potential. South Korea excels in cutting-edge semiconductor manufacturing, the widespread use of AI in daily life, and leading AI companies across various sectors—key factors that OpenAI is keen to collaborate with.
The open-source multi-agent development framework Rowboat, supported by the renowned Y Combinator incubator, has been released. The framework is designed to greatly simplify the creation of smart assistants. Rowboat supports multi-party collaboration platforms (MCP) and OpenAI’s Agent SDK, enabling developers to build complex agent workflows in just a few minutes.
Meta Establishes New AI Lab in Pursuit of “Superintelligence”
Meta is setting up a new AI research lab focused on achieving "superintelligence"—an AI system theoretically capable of surpassing human brain capabilities. This move is a key part of the company's AI strategy overhaul, with Meta inviting Alexandr Wang, founder of Scale AI, to join and planning a multi-billion-dollar investment in Scale AI. According to sources, Meta has offered compensation packages ranging from seven to nine figures to dozens of researchers from leading AI companies like OpenAI and Google, some of whom have already agreed to join.
LiquidStack Unveils GigaModular CDU to Tackle AI Data Center Cooling Challenges
LiquidStack has introduced its innovative GigaModular cooling liquid distribution unit (CDU), a modular and scalable solution designed to address the severe cooling challenges posed by artificial intelligence (AI) workloads in data centers.
The GigaModular CDU utilizes direct-chip liquid cooling technology, which is far more efficient than traditional air conditioning systems in handling the high heat density generated by AI hardware. Its modular design allows data center operators to scale cooling capacity from 2.5 MW to 10 MW or more, aligning cooling investments with actual heat load requirements and avoiding over-provisioning.