Today's AI News Roundup (August 4, 2026)
China's CXMT Closes In on Samsung and SK hynix in Cutting-Edge Memory
CXMT, China's largest state-backed memory chip maker, is on track to soon produce advanced low-power memory (LPDDR) for smartphones that matches the industry's state of the art. If prototype test results come back strong, the company will start small-volume production around the end of the year, with mass production to follow, Bloomberg reported, citing people familiar with the matter.
The company is on the Pentagon's blacklist, but the development shows that China is rapidly catching up with South Korea's Samsung Electronics and SK hynix in advanced memory, a field that is also indispensable for AI deployment. While CXMT still has some distance to cover before mass production, the speed at which the technology gap is narrowing is drawing renewed attention.
Huawei Scientist Points to the "Physical Limits" of Semiconductors and the Tau Scaling Law
There was another symbolic statement from a Chinese player. Liao Heng, Huawei's top semiconductor scientist, appeared in an interview of roughly four hours streamed in China in late July, arguing that the "physical limits" that major Western chipmakers such as NVIDIA face as they chase ever more powerful processors are drawing near, according to Bloomberg.
Reflecting on the company's experience of being cut off from global supply chains by U.S. sanctions, he claimed that a proprietary design approach Huawei has developed, which it calls the "Tau scaling law," works to its advantage. It is noteworthy that the argument — that going its own way amid continuing U.S.-China technology restrictions actually yields an edge — was made publicly. That said, technical verification remains limited, and the remarks appear to carry a promotional flavor as well.
AI Is Solving Tough Erdős Problems in Mathematics One After Another
Quanta Magazine published a detailed report on how the difficult problems left behind by the legendary mathematician Paul Erdős are being solved one after another with AI. AI-assisted proofs are reaching problems that had long gone untouched, suggesting that the very methodology of mathematical research is changing.
Around the "AI versus mathematics" debate, voices within the mathematics community itself are raising questions about the reliability of verification and the role of humans. Views are split between those who see these as landmark achievements and those who urge careful deliberation on how the verification process should be designed, and how proofs are evaluated looks set to become a major point of contention going forward.
Japan's Generative AI Usage Ranks Among the Lowest in the World, with Most Saying They Have "No Plans to Use It"
Some striking data has emerged for readers in Japan. According to an estimate published on Qiita, when Japan's generative AI usage rate is estimated as a share of the population, it sits at the bottom of the pack compared with countries such as the United States, India, China, and South Korea, with the majority saying they have "no plans to use it." The comparison covered the United States, Canada, Japan, South Korea, China, India, and Thailand.
Note that the estimate is based on interviews regarding ChatGPT and Copilot, so it is not rigorous statistics. Even so, the gap in adoption is consistent with reporting from other countries, and a fresh analysis of what is holding back usage in Japan seems warranted.
LinkedIn Won't Pour Big Money into AI Hardware This Year
There was also an unusual move amid the overheated AI infrastructure investment boom. LinkedIn has indicated that it will forgo large-scale spending on AI hardware this year, TechRadar reported. The specific reasons have not been covered in detail, but the company appears to be prioritizing capital efficiency and making the most of its existing infrastructure.
At a time when big tech companies are rushing headlong into massive investments, the decision to deliberately hold back on hardware spending stands in sharp contrast. It is a fresh reminder that not every player is pursuing capital investment at the same pace.