{"resolved":true,"match_type":"uuid","source_type":"original_book","book_id":"8775e0ee-45b3-49e0-9051-2909d31de8ea","title":"数字工厂","author":null,"edition":null,"publisher":null,"year":null,"isbn":"9787523600474","requested_by":"printed_page","book":{"source_type":"original_book","book_id":"8775e0ee-45b3-49e0-9051-2909d31de8ea","id":"8775e0ee-45b3-49e0-9051-2909d31de8ea","uuid":"8775e0ee-45b3-49e0-9051-2909d31de8ea","title":"数字工厂","normalized_title":"数字工厂","author":null,"edition":null,"publisher":null,"year":null,"isbn":"9787523600474","isbns":["9787523600474"],"parsed":true,"ai_readable":true,"document_type":"book","page_count":281,"navigation_url":"https://workshop.closeclaw.top/ai/book/8775e0ee-45b3-49e0-9051-2909d31de8ea","title_url":"https://workshop.closeclaw.top/ai/title/%E6%95%B0%E5%AD%97%E5%B7%A5%E5%8E%82","fixed_page_url_template":"https://workshop.closeclaw.top/ai/book/8775e0ee-45b3-49e0-9051-2909d31de8ea/page/{printed_page}","path_api_url_template":"https://workshop.closeclaw.top/api/ai/book/8775e0ee-45b3-49e0-9051-2909d31de8ea/page/{printed_page}","read_url":"https://workshop.closeclaw.top/api/ai/read/8775e0ee-45b3-49e0-9051-2909d31de8ea","page_url":"https://workshop.closeclaw.top/api/ai/books/8775e0ee-45b3-49e0-9051-2909d31de8ea/page","search_url":"https://workshop.closeclaw.top/api/ai/books/8775e0ee-45b3-49e0-9051-2909d31de8ea/search"},"page":{"printed_page":143,"pdf_page":147,"physical_index":147,"physical_index_base":0,"printed_page_status":"recognized","page_sha256":"25c917cdc0ba8a48993c0b623a2006a8398dd85f9f65b8ca52948002a25a686f"},"chapter":"第3章分散的工厂：众包工作","content":"由于安全至关重要，无人驾驶汽车必须尽可能地识别出自己所遇到的一切，包括其他车辆、行人、骑自行车的人、红绿灯、交警检查岗、动物、建筑工地和坑洼等。为了训练算法，开发者需要大量带有注释的照片和视频素材。这就是为什么全球会有那么多工人登录众包工作平台，花费大量的时间在视频和图片素材中标记、注释物体。这些素材大多是车辆行驶过程中拍摄的视频。工人的任务包括进行各种形式的标记，并指出镜头中的不同对象。其他工人会复核同事所作的标记，或者各种算法在模拟过程中作出的判断。在训练数据以及精度方面的新要求，不仅导致许多现有的平台专注于为机器学习进行数据注释、还推动创建了专门满足自动驾驶汽车软件开发者需求的新平台（比如 Scale 和Mighty AI)。①Mighty AI平台（其面向工人的界面被称为Spare5）于2019年被优步收购，关闭了面向其他客户的业务，专门为优步的自动驾驶汽车开发服务。\n\n众包工作平台的外包劳动力在许多领域都发挥着重要的作用。与之相似，众包工人对自动驾驶系统所做的隐形贡献也说明，备受瞩目的自动化过程是如何由人力大举推动的。数十万平台工人注释数据的“幕后工作”对自动驾驶汽车的发展至关重要。①和自动化方面的许多问题一样，围绕自动驾驶讨论的论文总是专注于技术，往往夸大了算法的能力，却忽略了自动化过程中所需的人力劳动。\n\n1第3章1分散的工厂：众包工作\n\n①另参见 Schmidt, Crowdproduktion von Trainingsdaten.\n\n143","previous_page":142,"next_page":144}