东京 —— 日本人工智能初创公司 EpicAI 计划向国内汽车制造商部署其软件，以加速车辆设计评审，因为这些车企难以跟上中国电动汽车竞争对手以快得多的速度推出新车型的步伐。
TOKYO -- Japanese artificial intelligence startup EpicAI plans to deploy its software to accelerate vehicle design reviews at domestic automakers, as they struggle to keep pace with Chinese electric vehicle rivals bringing new models to market at a much faster pace.
该公司的软件有望大幅缩短设计新车所需的时间和人力，旨在利用日本传统上谨慎的制造商在竞争加剧之际寻求降低成本途径的心态转变。
The company's software, which could dramatically reduce the time and manpower needed for designing a new vehicle, aims to take advantage of a shift in mindset among Japan's traditionally cautious manufacturers seeking ways to reduce costs as competition intensifies.
EpicAI 是一家从东京科学大学分拆出来的公司，计划从明年春天开始在几家日本汽车制造商处引入其软件，此前的试用据称获得了客户的积极反馈。
EpicAI, a spin-off from Institute of Science Tokyo, plans to introduce its software at several Japanese automakers from next spring, following trials that it says have received positive feedback from customers.
“以将设计流程所需时间减半为目标，这绝非易事，”EpicAI 首席执行官横山惠一在最近接受《日经亚洲》采访时表示，并援引了日本车企激进的车辆开发精简计划。
"Targeting, for example, halving the time needed for the design process is no easy feat," Keiichi Yokoyama, chief executive of EpicAI, told Nikkei Asia in a recent interview, citing Japanese carmakers' aggressive plans to streamline vehicle development.
他补充道：“要解决这些问题，必须使用人工智能等根本性的创新工具。”
"To tackle those problems, you have to use fundamentally innovative tools such as AI," he added.
这家总部位于东京的初创公司成立于 2023 年，目前拥有约 50 名员工。其业务涵盖 AI 咨询和定制 AI 解决方案，客户包括汽车零部件制造商电装和特斯拉供应商松下等日本大型企业。
The Tokyo-based startup was founded in 2023 and now has about 50 employees. It engages in businesses such as AI consulting and custom AI solutions, and its clients include major Japanese companies including auto parts maker Denso and Tesla supplier Panasonic.
横山惠一曾是与研究员松尾丰相关的知名 AI 实验室的成员。松尾丰一直是塑造日本 AI 政策的主要声音之一，担任政府 AI 战略专家会议主席。
Yokoyama is a former member of the prominent AI laboratory associated with researcher Yutaka Matsuo. Matsuo has been one of the leading voices in shaping Japan's AI policies, assuming the role of chair of the government's AI Strategy Expert Panel.
日本初创公司EpicAI首席执行官横山圭一表示，日本汽车制造商可以利用由人工智能驱动的创新工具，精简新车型开发流程。（Epic AI）EpicAI的技术旨在实现审查车辆设计潜在故障这一繁琐过程的部分自动化，该过程长期以来是日本车企安全可靠声誉的核心，但越来越被视为加快开发速度的障碍。
Keiichi Yokoyama, chief executive of Japanese startup EpicAI, says that Japanese automakers can streamline new model development with innovative tools driven by AI. (Epic AI) EpicAI's technology is designed to automate parts of the painstaking process of reviewing vehicle designs for potential failures, which has long been central to Japanese automakers' reputation for safety and reliability but is increasingly seen as an obstacle to faster development.
该公司瞄准的流程被称为“失效模式与影响分析”（FMEA），工程师在其中检查零部件或系统可能如何失效，并评估潜在后果。横山表示，这是质量控制的关键环节，因为许多车辆召回源于设计阶段未能预见潜在失效模式，而非制造缺陷。
The company is targeting a process known as Failure Mode and Effects Analysis, or FMEA, in which engineers examine how a component or system might fail and assess the potential consequences. This is a critical part of quality control, as many vehicle recalls stem not from manufacturing defects but rather from failing to anticipate potential failure modes during design, Yokoyama said.
虽然该流程在全球汽车行业广泛应用，但丰田汽车等日本制造商传统上尤为重视详尽的质量检查。
While the process is widely used across the global automotive industry, Japanese manufacturers such as Toyota Motor have traditionally placed particular emphasis on exhaustive quality checks.
质量既是日本公司备受推崇的原因，也是令人担忧的来源。2024年，汽车行业的多家日本企业面临安全测试丑闻和违规问题。
Quality is both a reason Japanese companies are held in high regard and a source of concern. A number of them in the automotive industry faced a safety test scandal and irregularities in 2024.
据横山介绍，审查单个零部件的设计可能需要一名工程师耗费约200小时，这使其成为车辆开发中极为耗时的环节。
According to Yokoyama, reviewing the design of a single component can require roughly 200 hours of work by a single engineer, making it a time-consuming part of vehicle development.
因此，EpicAI的目标是将审查车辆设计的时间从约200小时大幅缩减至仅几小时。
EpicAI is therefore targeting a dramatic reduction in the time spent reviewing vehicle designs, from roughly 200 hours to just a few hours.
其软件利用多个生成式AI模型协同工作，分别承担分析零部件规格、识别潜在风险及评估可能的失效场景、交叉验证彼此发现以减少错误等不同任务。
Its software uses multiple generative AI models that work together on different tasks such as analyzing component specifications, identifying potential risks and assessing possible failure scenarios, and cross-checking one another's findings to reduce errors.
横山表示，该软件将首先在多家日本车企的部分部门部署，更广泛的应用将视进一步讨论而定。他拒绝透露涉及的公司名称。
Yokoyama said the software would initially be deployed in selected departments at multiple Japanese automakers, with broader adoption subject to further discussions. He declined to identify the companies involved.
EpicAI迎来商机之际，丰田、日产汽车和本田汽车等老牌车企正寻求大幅缩短新车型开发时间。
Business opportunities for EpicAI come as established carmakers, including Toyota, Nissan Motor and Honda Motor, seek to drastically shorten the time needed to develop new models.
本田作出了最激进的承诺之一，计划从2028年起将开发时间减半；日产则表示，计划于今年发布的新一代Skyline车型，开发周期有望缩短至26个月，不足以往的一半。过去，开发一款新车型通常需要约四年时间。
While Honda has made one of the most aggressive pledges to halve development time starting from 2028, Nissan said the next version of its Skyline model, slated to be unveiled this year, is likely to be developed in 26 months, less than half the usual time. In the past, it was common practice to spend about four years developing a new model.
本田已与Rescale建立合作伙伴关系，后者是一家美国研发平台提供商，获得了英伟达和OpenAI创始人萨姆·奥特曼的支持。日产则与英国Monolith等AI工具开发商合作，利用AI精简设计和测试流程。
Honda has partnered with Rescale, a U.S. development and research platform provider backed by Nvidia and OpenAI founder Sam Altman. Nissan, meanwhile, has partnered with AI tool developers such as the U.K.'s Monolith to streamline the design and testing process with AI.
由于中国竞争对手的激烈竞争，这些车企面临巨大压力，中国竞争对手正逐渐在中国和东南亚市场蚕食日本车企的份额。
The automakers are under pressure from fierce competition with Chinese rivals, which are gradually winning market share in China and Southeast Asia at the expense of their Japanese rivals.
中国电动汽车制造商以极快的速度开发新车型著称。例如，电动汽车巨头比亚迪用约两年时间开发了其首款微型车Raccoon，尽管日本独特的轻型车（kei car）细分市场对该车企而言是全新领域。
Chinese EV makers are well known for developing new vehicles at a rapid pace. For example, EV giant BYD developed its first minicar, the Racco over approximately two years, even though Japan's unique kei car category was a completely new segment for the automaker.
伊藤忠综合研究所执行研究员深尾三四郎表示，中国电动汽车制造商通过接连推出新车型积累经验，从而生产出更具吸引力的产品。他说：“要与中国竞争对手抗衡，日本车企比以往任何时候都更需要专注于强化品牌。”“AI是手段，而非目的本身。”
Sanshiro Fukao, executive fellow at Itochu Research Institute, said Chinese EV manufacturers are gaining experience by launching new models in rapid succession, which has led them to produce more appealing products. "To compete with Chinese rivals, Japanese automakers need to focus more than ever on strengthening their brands," he said. "AI is a means to an end, not an end in itself."