伪娘

经纬度：中国AI从“跟跑”加速迈向“领跑”_我的网站

A |     7月20日上午，在国务院新闻办公室举行的新闻发布会上，工信部介绍，今年以来深化拓展“人工智能+”，规模以上制造业企业人工智能技术应用普及率超30%，人工智能开源大模型全球累计下载量突破100亿次。         100亿次的下载量，不仅是一个里程碑式的数据，更是一个清晰的信号——中国的人工智能产业，正在从“跟跑”加速迈向“领跑”的新阶段。    

    The USS Gerald R. Ford departs from Souda Naval Base near Chania on the island of Crete, Greece, on February 26, 2026. Photo: VCGUS President Donald Trump's reported push to reshape the US Navy's next-generation aircraft carriers - from moving the island to abandoning electromagnetic catapults in favor of steam-powered systems - could impose heavy engineering, financial and schedule costs, Chinese military affairs experts said, warning of a widening gap between political preferences and the realities of building some of the world's most complex warships.
The Washington Post reported on Saturday that the US Navy is considering a major redesign of its new Ford-class aircraft carriers to better match Trump's preferred aesthetic, with the president favoring ships that more closely resemble US carriers from the World War II era. One option under study would move the carrier's island farther forward, toward the middle of the ship.
US officials cited by the report warned that such a change could require billions of dollars in redesign costs and cause substantial delays. 
The potential island redesign comes only days after Trump ordered another far-reaching change to the Ford-class program. On Thursday, he signed a national security memorandum directing the Pentagon and Navy to develop a plan within 60 days to replace the Electromagnetic Aircraft Launch System (EMALS) with steam catapults and the Ford-class carriers' Advanced Weapons Elevators with hydraulic systems, beginning with the future USS Doris Miller, or CVN-81, USNI News reported. 
Trump has repeatedly criticized EMALS as "too complex" and "not as effective as steam catapults," according to Reuters. 
But the order would reverse a technology path the US Navy has pursued for more than a decade. When the Navy selected EMALS in 2009, it said the system would lower overall lifecycle costs, require less maintenance than steam catapults and place less physical stress on carrier-based aircraft, per Reuters.
As of March 2026, the USS Gerald R. Ford had conducted 36,863 EMALS launches, according to Pentagon documents cited by Reuters. 
General Atomics, the manufacturer of EMALS, has warned against changing course on the Doris Miller, saying nearly 50 percent of production related to the system has already been completed and that a switch now would create significant cost, schedule and integration risks, according to Reuters. 
Zhang Junshe, a Chinese military affairs expert, told the Global Times on Sunday that the Ford-class carrier's aft-positioned island is not merely an aesthetic choice, but is closely tied to flight-deck efficiency and landing safety. "Moving the island farther aft creates more space around the catapults for aircraft parking and handling, helping shorten launch preparation time, while also reducing airflow disturbance during aircraft recovery," he said. 
Relocating the island forward would not be a simple cosmetic adjustment, but a major redesign affecting weight distribution, buoyancy, flight-deck layout and aerodynamics, with significant implications for both cost and schedule, Zhang noted.
Song Zhongping, another Chinese military affairs expert, told the Global Times that Trump's demand is one thing, but turning it into an actual warship design is another.
"The first four Ford-class carriers have either entered service or are already in different stages of construction, leaving very little room for fundamental changes to their overall designs," Song said. Even the fifth and sixth ships - the future USS William J. Clinton, CVN-82, and USS George W. Bush, CVN-83 - would face considerable cost and engineering consequences if the Navy attempts to overturn their established technological architecture, he said.
The US Navy officially announced the names of CVN-82 and CVN-83 in January 2025, per the US Navy official website. 
"There is an obvious gap between US President's political will and the engineering realities of aircraft-carrier construction," Song said.
Such proposed changes come as Ford-class construction timelines have already grown considerably compared with those of the preceding Nimitz class. The US Congressional Budget Office (CBO) said in April that Nimitz-class carriers typically took seven to eight years to build, while Ford-class ships are currently taking roughly 10 to 11 years. In a separate assessment released this month, the CBO estimated that future Ford-class carriers under the Navy's 2027 shipbuilding plan could cost about $22 billion apiece. 
"Electromagnetic launch represents an advanced direction in aircraft-carrier catapult technology and has far greater development potential than traditional steam catapults," Song said. "If the US gives up electromagnetic catapults and returns to steam catapults, that would be a technological regression and a step backward."
Echoing Song, Zhang said electromagnetic catapults offer more precise control of launch energy and broader compatibility with different types of carrier-based aircraft, while requiring less shipboard space than steam systems with their extensive high-pressure piping.
"Replacing EMALS with steam catapults on a Ford-class carrier already designed around electromagnetic launch would not only represent a technological step backward, but could also turn an attempt to reduce risk into a new source of engineering and system-integration risk," Zhang said.
。         7月17日，在2026世界人工智能大会开幕前夕，月之暗面发布Kimi K3大模型。视觉中国 图          国产AI行业厚积薄发          近年来，AI浪潮席卷而来，一场如火如荼的AI竞赛，正在全球上演。

B | 在如此激烈的竞争环境下，一批国产AI产品接连横空出世，走到世界舞台的中央。比如早期的引发全行业关注的“Deepseek时刻”；再如前几天，月之暗面发布的Kimi K3，成为全球首个开源的3万亿级大模型，再次引发了全球科技圈的热议。         这些大模型产品的密集涌现，不是偶然的单点突破，而是国产AI行业厚积薄发的结果。         诚然，中国互联网用户规模大，企业数量多，有市场基数的规模优势。但高达100亿次的下载量，显然无法用人口红利来简单解释。事实上，从硅谷的互联网平台，到欧洲的传统制造工厂，越来越多的海外机构、企业，都纷纷主动选择国产大模型，作为其技术底座。         这一方面是因为，国产大模型通过持续的技术创新，以及积极的开源策略，把价格“打”下来了，降低了使用成本；另一方面，国产大模型在参数、性能、稳定性等各方面，已经建立了独特的竞争优势。         国产AI的这种技术进步，相信普通人也有直观的感受。一两年前，AI大模型之于普通人，更多还是“聊天机器人”的概念，再就是能撰写文案、做出漂亮的PPT，更像是一个提升办公效率的辅助工具。但如今AI的技术演进，已经远远超出了这一范畴。

C |          我国规模以上制造业企业普及率超30%的背后，就离不开大模型的支撑，AI开始深入流水线，参与工业质检、流程优化等各个环节。此外，像参与智慧城市的调度管理，甚至进医院帮助医生辅助诊疗等，都成了大模型落地的确定性场景。目前，人工智能不仅创造了1.2万亿的产业规模，还正在赋能千行百业，成为实体经济转型的重要推动力量。         中国的工业门类齐全，制造业场景丰富，为AI的应用提供了良好的土壤。

D | 而大模型的广泛落地，又为AI提供了海量、真实的数据，能够持续反哺大模型的迭代与优化，形成良性循环，为AI产业注入源源不断的竞争力。         是里程碑，更是新起点          当然，在肯定成绩的同时，我们也需保持清醒的认知。人工智能不同于传统产业，它是一个“一日千里”的行业，今天领先的技术路线，过几个月可能也会被新的方案取代。要巩固当前的技术优势，就要保持危机意识，不断推动底层架构和核心算法的创新。         像这次发布会就提到，推动建立算力服务能力评估、算力市场化定价等标准。这也是完善AI产业基础设施的重要一步。         算力被视作人工智能行业的“水电煤”，是支撑创新的基础设施，建立市场化的定价标准，让算力像“水电煤”明码标价、按需分配，可以确保算力的供应更加透明、公平，降低中小企业使用AI的门槛，进一步推动产业发展。         另外，发布会还提到，东部、中部、西部、东北地区智算规模占全国比例，分别是55.9%、10.6%、32.6%、0.9%。

E | 这一数据也意味着，AI算力资源的分布，仍然存在着较大的地区差距，未来要进一步平衡发展落差，补齐落后区域的短板。         说到底，人工智能发展的最终目的，不是追求领先的参数、实现惊人的下载量，或者让国产AI产品在排行榜上拿到好看的名次，而是让这项技术真正普惠到全国各地，到各行各业，让每个人都能从中受益。         100亿次下载量，是一个里程碑，更是一个新起点。         未来，在加快技术创新的同时，还要进一步推动AI的应用和普及，不断降低使用门槛。只有让AI像“水电煤”一样，真正进入社会的每一个角落，让普通人“用得上”也能“用得起”好的AI产品，中国的人工智能产业革命才能释放最大的价值。         （“经纬度”系澎湃新闻打造的经济评论IP，用主流媒体新表达助力中国经济话语和中国经济叙事体系建设。）          海报设计 王璐瑶。

Current article:http://www.shangfochengcongshuaizhuang.buzz/vcckh/1wa.html

Published on:00:05:28