Alphabet旗下的人行道实验室（Sidewalk Labs）推出了一种名为“Delve”的工具，可利用人工智能在几分钟内为城市创造“数百万种设计可能性”。该工具可根据预算、位置和大小等标准创建候选方案，然后对它们进行排名，以便开发者选择最好的设计。Alphabet subsidiary Sidewalk Labs has launched Delve, a tool that uses artificial intelligence to create "millions of design possibilities" for urban developments in minutes. Delve creates options based on criteria such as budget, location and size and then ranks them so developers can choose the best design.该项目采用了机器学习（machine learning）。这是一种人工智能应用，使用一组基本数据，根据经验进行学习和改进。为了建立深入设计的基础，实验室制作了一个核心组件的初始模型，这一组件通常用于社区开发。The project leverages machine learning, an application of artificial intelligence that uses a base set of data and learns and improves with experience. To create the basis for Delve designs, Sidewalks Labs made a starter model of core components typically used in neighbourhood developments.△ Delve演示页面 图源：Dezeen 实验室的产品管理总监Okalo Ikhena说道：“城市社区有独特的个性，但他们也有很多相同的核心组成部分。”“我们的团队已经建立了这些核心组件的模型，包括建筑、开放空间、便利设施、街道和能源基础设施。”"City neighbourhoods have unique personalities, but they also share a lot of the same core components," Sidewalk Labs' director of product management Okalo Ikhena told Dezeen. "Our team has built a model of these core components that includes buildings, open spaces, amenities, streets, and energy infrastructure," Ikhena added.在流程开始时，客户添加项目信息，包括规划数据，如商业和住宅专用面积；用地限制，例如建筑物高度限制；以及成果优先级，包括期望的成本和日照。一旦这项工作完成，该工具就会产生数以百万计的候选方案，包含了每个方案如何满足客户的关键优先级的信息，再根据这些信息对方案进行评级。At the start of the process, customers add project information including the planning input, such as the area dedicated to commercial and residential space; site constraints, such as building height limits; and the priority outcomes, including the desired cost and amount of daylight. Once this is completed Delve generates millions of options with information on how each meets the customer's key priorities – it grades the proposals according to this, so the user can choose the most successful option.
△ Delve演示页面 图源：Dezeen Ikhena称，Delve之所以能脱颖而出，是因为它提供关于“整体”要素的信息，比如便利设施的步行可达性、户外空间和日照。它评估了单个住宅单元接收到的日照，而不仅仅是整个建筑，建筑师从而可以相应地选择材料。According to Ikhena, Delve stands out because it provides information on "holistic" elements like walkability to amenities, outdoor space and daylight. It assesses the amount of daylight that individual residential units receive, rather than the whole building, so that materials can be chosen accordingly.“其他产品只关注一个或几个指标——它们不评估生活的整体质量。”“我们认为，如果不考虑如何达成整体的成功（日照、步行性、景观、成本），一个区域的设计最终是不会成功的。”"Other products only focus on one or a few metrics – they don't evaluate overall quality of life," he said. "We think designing a district without considering how it will be holistically successful (for daylight, walkability, views, cost) ultimately will not be successful."设计师可以在整个开发过程的不同阶段引入该工具，从寻址阶段到可行性评估阶段，都可创建现有模型的替代品。内置新型能源需求预测系统，以确保开发商能够满足能源效率和可持续发展的目标。In addition to developing designs from scratch, Delve can be introduced at different stages throughout the development process, from scoping out ideal sites through feasibility assessments to creating alternatives to an existing model. Customers are given a built-in utility model which forecasts energy requirements. This constantly updates so developers can ensure they meet energy efficiency and sustainability goals.
△ Delve演示页面 - 方案对比 图源：Dezeen 开发商Quintain最近在伦敦附近的温布利公园为一个12英亩的多功能开发项目设计了一个可租赁建筑方案。该工具提供了24个高性能的方案，超出了开发商在采光、室外空间和住房单元方面的预期。Developer Quintain recently used Delve to design a Build to Rent scheme on a 12-acre mixed-use development in Wembley Park, near London. The tool created 24 high-performing schemes that exceeded the developer's priorities on daylight, outside space and housing units.有人预测，机器学习在设计过程中的兴起，将导致90%的建筑师失业。然而，对于如何利用机器学习来帮助建筑师和设计师的工作，Ikhena持更为积极的观点：“该工具旨在提升用户（设计师）的能力。它可以将手工流程自动化，促进人们将注意力集中在关键战略决策上。”At the time Errazuriz predicted the rise of machine learning in the design process would result in 90 per cent of architects losing their jobs. Ikhena, however, has a more positive view on how machine learning can be used to aid the work of architects and designers. "The tool is designed to greatly improve the capabilities of users by automating manual processes (counting apartments of producing area takeoffs) and focusing attention on key strategic decisions (which circulation approach or program mix to test)."