Apple confirms that the US version of iPhone 18 Pro Max has problems; OpenAI spends more than $500,000 a day to investigate AI intrusion incidents; job seekers complained about the "uncanny valley" of AI interviews and became a hot search | Geeks Already Known苹果确认美版 iPhone 18 Pro Max 有问题；OpenAI 每天烧超 50 万美元查 AI 入侵事故；求职者吐槽 AI 面试「恐怖谷」上热搜｜极客早知道

Apple confirmed: There is a problem with the US version of iPhone 18 Pro Max. On October 3, Beijing time, Bloomberg reported that Apple said that problems with AT&T have caused some iPhone 18 Pro Max users to be unable to make calls, access data, or send and receive text messages while online. It's unclear how many people are experiencing this issue. Affected users found their devices largely unusable unless connected to a Wi-Fi network. Apple has released a carrier settings update and rolled out a new version 27.0.1 of the iOS operating system to prevent the issue from spreading further. The carrier settings fix will be automatically downloaded for AT&T customers over the next 14 days, but Apple still recommends consumers install any available software updates immediately. An Apple spokesperson said in a statement provided to Bloomberg: "We have discovered a glitch affecting a small number of iPhone 18 Pro Max users on the AT&T network that may cause the device to lose service and make calls." According to the report, for users who have been affected, the only solution is to replace the phone and use a new device provided by Apple or AT&T. The report mentioned that the iPhone 18 Pro Max sold in the United States is a little different from other new models: it is equipped with a cellular baseband manufactured by Qualcomm. All other versions, including the smaller iPhone 18 Pro, are equipped with Apple’s own C2 baseband. According to user reports, only the iPhone 18 Pro Max is affected by AT&T's vulnerability. (Source: Sina Technology) NVIDIA Vera Rubin NVL72 ran 4.8 times Token throughput AI cloud service provider CoreWeave recently announced that the NVIDIA Vera Rubin NVL72 system has been officially put into production on CoreWeave Cloud, and AI software company Cognition has become the first customer to run production workloads on the platform. Cognition runs the SWE-2 inference task on Vera Rubin NVL72, and the measured total token throughput is up to 4.8 times higher than that of GB200 NVL72. In reinforcement learning scenarios, the throughput per GPU output token also reaches 3.8 times that of GB200 NVL72, which means Devin can handle more sessions simultaneously and accelerate multi-step reasoning. In the DeepSeek-R1 inference benchmark test, Vera Rubin NVL72's token throughput per megawatt reached 10 times that of GB200 NVL72, and can produce more intelligent output under the same power consumption. In terms of hardware specifications, the NVL72 system is composed of 36 Vera CPUs and 72 Rubin GPUs, and adopts a 100% liquid-cooled design. The cable-free modular tray design reduces installation time from two hours to five minutes, greatly improving deployment efficiency. (Source: Kuai Technology) Its agents frequently get into trouble, and OpenAI spends more than $500,000 a day to investigate. According to the British "Guardian" report on the 3rd, OpenAI disclosed that in order to investigate incidents such as its AI agents attacking the Australian Medicare medical insurance system and Hugging Face, the company invested more than $500,000 a day (approximately 3.357 million yuan at the current exchange rate). Due to the sheer size of the data that needs to be screened, OpenAI also uses AI to participate in the review. OpenAI blogged this week about the scale of the investigation: The company needed to examine 50 petabytes of data, or about 50 million gigabytes. If all this data were given to one person to read, it would take about 66 million years. OpenAI warned that the investigation is not over yet, and more institutions may be notified in the near future that they have been targeted by the agent's actions. The review focused on records of models accessing or modifying websites, as well as actions involving passwords, application programming interface access and other sensitive credentials. OpenAI uses AI to help sift through this data, investing more than $500,000 a day