从古代对人类智力的哲学探索到前沿创新，人们对创造智能的追求跨越了时代。如今，人工智能（AI）正迅速成为当今社会的一个普遍现象，加州伯克利大学的斯图尔特-拉塞尔等人工智能科学家预测，人工智能将成为未来的主导技术。这一发展的时间表和进展情况很难准确预测，但我们必须为机器在实际工作中远超人类决策能力的可能性做好准备。获得更高的智能将是人类历史上最大的事件。人工智能科学家斯图尔特认为，重要的是要明白 "这也可能是人类历史上最后一件大事，以及如何确保它不会发生。" [ Stuart Russel (2023) Human Compatible.人工智能与控制问题。]
From ancient philosophical inquiries on human intellect to cutting-edge innovations, the quest to create intelligence has spanned the ages. Today, Artificial Intelligence (AI) is rapidly becoming a pervasive aspect of the present and it is predicted by AI scientists like Stuart Russel of University of Berkeley, California to become the dominant technology of the future. The timeline and progress of this development is difficult to predict with certainty but we must plan for the possibility that machines will far exceed the human capacity for decision making in the real work. Gaining access to considerably greater intelligence would be the biggest event in human history. According to the AI scientist Stuart it is important to understand that ‘It also might be the last event in human history and how to make sure that it is not.’
1
人工智能在城市规划中的应用——情况报告
近年来，人工智能（AI）在城市规划中的应用成为众多规划专家和机构讨论和研究的主题。2022 年，联合国人居署分析了 "人工智能与城市 "的风险、应用和治理。联合国人居署指出，"人工智能领域正在以一种肆无忌惮的速度发展。我们越来越多地看到人工智能系统离开研究环境，部署到人类活动的几乎所有领域。因此，人工智能有可能深刻改变我们社会的运作方式，包括支持解决气候危机、公共卫生、教育等关键问题的努力。然而，这种持续的社会变革也带来了必须应对的风险。目前迫切需要在公共和私营部门的所有规模级别的行政和政治组织中发展负责任的人工智能治理和实践"。
联合国人居署的报告为如何在城市和住区背景下负责任地部署人工智能提供了第一个总体框架。它是联合国人居署指导地方当局在其城市和住区实现以人为本的数字化转型进程的战略的一部分。在此过程中，必须考虑自 2022 年以来出版的其他相关文献。
例如，卡拉尔顿市首席信息官克里斯-钱科内（Chris Chiancone）在 2023 年 6 月 20 日发表的论文《利用生成式人工智能革新城市规划：智能城市的新时代》中写道："生成式人工智能的核心是一种机器学习，它可以根据训练集制作新的数据实例。想象一下，一位艺术大师不是简单地复制一个场景，而是运用自己的创造力，从给定的输入中产生独特、逼真的输出。这就是生成式人工智能的神奇之处，但却是在数据领域。它是数字世界的艺术大师，挥舞着算法和计算能力，而不是画笔和颜料"。
城市规划是一个多层面的领域，需要对大量数据进行整合和审查。传统上，城市规划者必须手动筛选这些数据，并对未来趋势做出明智的预测。这种方法不仅耗费大量人力，而且容易出现人为错误和偏差。生成式人工智能在分析大型数据集和绘制逼真的城市蓝图方面表现出色，为克服这些障碍架起了一座桥梁。此外，生成式人工智能还能预测城市扩张并优化基础设施规划。
InfiniCity的街道是一个利用人工智能建造的三维城市合成模型，其目标是 "自动再现栩栩如生的现实世界城市"，完全由合成头脑的想象力生成。彭博社论文《人工智能设计城市时的样子》（What it looks like When AI Designs a City）的作者帕特里克-西瑞（Patrick Sission）认为，这就像是一篇由 ChatGPT 撰写的学期论文的三维视图，陷入了胡编乱造的引文和笨拙的语法。现实世界中城市的有机逻辑还不能完全归结为模拟城市式的平台。尽管 InifiniCity 看起来令人着迷，但在那里生活并没有太大的吸引力。在网上可以找到一些通过 Midjourney 等程序创建的虚构案例。与此同时，城市官员也开始思考人工智能在城市治理中的相关性和风险。
简而言之，对于Chiancone 来说，生成式人工智能为城市规划带来了创造力与效率的独特融合。这就好比有了一个超级助手，可以处理数字、分析数据、生成模型，并以人类无法比拟的速度和规模进行预测。所有这些都是为了创造更高效、更宜居和更可持续的城市。
生成式人工智能与城市规划的交叉点是技术与创意的奇妙融合。其中存在风险，包括道德和数据隐私方面的考虑，也存在技术限制。确保公平结果、保护个人隐私和克服技术限制都是至关重要的任务，需要仔细关注和深思熟虑的解决方案。如果管理得当，它将为管理城市环境的复杂性提供有力的工具，使城市更智能、更可持续、更公平，并带来高品质的生活。利用数字技术提高绩效和福祉的城市可以为创建 "智慧城市 "铺平道路。
迄今为止，人工智能的发展有助于创造印象派图像和艺术作品，为想象力提供便利，但它并不能生成蓝图。帕特里克-西瑞特（Patrick Sission）介绍说，扎克-卡茨（Zach Katz）利用人工智能技术创造了更多乌托邦式的城市，他并不太担心人工智能可能会被滥用；在他看来，人工智能辅助城市图像的爆炸式增长反映了一种更健康的现象--人们对更美好的地方充满向往，热切希望看到城市被赋予更多智能。
卡尔斯鲁厄理工学院的 Daniel Podrasa、Peter Zeile 和 Markus Nepp 在他们的论文 "用于土地利用方案和城市设计的机器学习 "中进行了分析。他们讨论了有监督和无监督的机器学习算法，并得出结论：他们以人工神经网络分类为例讨论了有监督的学习方法，这种方法可以生成土地使用方案，并将其作为生成式城市设计的设计基础。然而，人工神经网络的训练和为生成训练数据而进行的土地利用描述仍然是一件相对耗时的事情。
"所提出的概念并不是要取代实际的规划工作，即计划的制定和计划的实现；相反，我们认为它是一种有益的补充，能够更快、更好、更透明地 "探索 "变体，不仅能够更好地掌握定性设计工作，而且能够更好地掌握定量城市发展设计。合理使用人工智能，有助于在早期阶段快速研究各种变体。人工智能的目的不是取代规划师的职业，而是更好地为决策奠定基础"。关于在城市治理中使用生成式人工智能的更具体分析，请参见 Nelson, RJ.Naing, KM.Beroche, H. 2023.生成式人工智能在城市治理中的应用》。
The use of Ar+ficial Intelligence in Urban Planning – a situa+on report
In recent years, the use of ArMficial Intelligence (AI) in Urban Planning became a subject of discussion and research by numerous planning experts and insMtuMons.2 In 2022, UN-Habitat analyzed the Risks,ApplicaMons and Governance of “AI and CiMes”3. According to UN-Habitat “The field of AI is growing at an unbridled pace. We are increasingly seeing AI systems leaving research se>ngs to be deployed in almost all spheres of human ac?vity. As a result, AI has the poten?al to profoundly transform the way our socie?es operate, including by suppor?ng efforts on cri?cal ques?ons such as the climate crisis, public health, educa?on and beyond. However, this ongoing societal transforma?on entails risks that must be addressed. There is an urgent need to develop responsible AI governance and prac?ces across all scale levels of administra?ve and poli?cal organisa?ons, in both the public and private sectors.”
The UN-Habitat report provides a first general framework on how to deploy AI responsibly in the context of ciMes and seXlements. It is part of UN-Habitat’s strategy for guiding local authoriMes in realizing a people-centered digital transformaMon process in their ciMes and seXlements. In doing so, other relevant literature has to be considered which has been published since 2022.
For instance, Chris Chiancone, Chief Information Officer at the City of Carralton writes in his paper of 20 June 2023 ‘Revolutionizing Urban Planning with Generative AI: A new Era of Smart Cities’4: “At its heart, Generative AI is a type of machine learning that crafts new data instances reflecting its training set. Picture a master artist who, rather than simply duplicating a scene, employs their creativity to produce unique, lifelike outputs from a given input. That's the magic of Generative AI, but in the realm of data. It's the digital world's virtuoso, wielding algorithms and computational power instead of brushes and pigments”.
Urban planning is a multifaceted field demanding the integration and scrutiny of a vast array of data. Traditionally, urban planners have had to manually sift through this data and make informed forecasts about future trends. This method is not only labor-intensive but also susceptible to human error and bias. GeneraMve AI, with its prowess in analyzing large datasets and cra]ing realisMc urban blueprints, provides a bridge to these hurdles. Furthermore, GeneraMve AI can forecast urban expansion and opMmize infrastructure planning.
The streets of InfiniCity, a model 3D city synthesis built using AI has as a goal to “automatically recreate lifelike real-world cities”, generated entirely from the imagination of a synthetic mind.5 According to Patrick Sission the author of the Bloomberg paper “What it looks like When AI Designs a City” 6 it’s sort of a 3D vision of a ChatGPT-authored term paper that descends into made-up quotations and awkward grammar. The organic logic of real-world cities can’t quite be reduced to a SimCity-esque platform just yet. As mesmerizing as InifiniCity looks it is not very much appealing to live there. Several fictional examples created via programs like Midjourney can be found online. Meanwhile, city officials are beginning to wrestle with the relevance — and risk — of AI in urban governance.
In a nutshell, for Chiancone Generative AI brings a unique fusion of creativity and efficiency to urban planning. It's akin to having a supercharged assistant that can process numbers, analyze data, generate models, and make predictions at a pace and scale that humans simply can't compete with. All of this is done with the aim of creating more efficient, habitable, and sustainable cities.
The intersection of Generative AI and urban planning is a fascinating blend of technology and creativity. There are risks including ethical and data privacy considerations and there are technical limitations. Ensuring equitable outcomes, protecting individual privacy, and overcoming technical limitations are crucial tasks that require careful attention and thoughtful solutions. If well governed, it offers a potent tool for managing the complexity of urban environments, leading to smarter, more sustainable, and more equitable cities with a high quality of life. Cities that use digital technologies to enhance performance and wellbeing can pave their way for the creation of ‘smart cities’.
AI as developed until now is useful to create impressionistic image and artwork to facility imagination but it doesn’t generate blueprints. Patrick Sission describes that Zach Katz, who has deployed the technology to create more utopian versions of cities, is less concerned about the potential misuse of AI; to him, the explosion of AI-aided urban imagery reflects something healthier — an appetite for better places, and an avid desire to see cities remade with a little more intelligence.
Daniel Podrasa, Peter Zeile and Markus Nepp of the Karlsruhe Institute of Technology analyze in their paper “Machine learning for land Use Scenarios and Urban Design”.7 They discuss supervised and unsupervised machine learning algorithms and conclude that a supervised learning method, which they discussed with the example of ArMficial Neural Network for classificaMon, land-use scenarios can be generated and these can be used as a design basis for GeneraMve urban design. However, the training of the ArMficial Neural Network and the descripMon for the land use for generaMng the training data is sMll a relaMvely Mme-consuming matter.
“The proposed concept is not intended to replace the actual planning work in the sense of plan elabora?on and plan realiza?on; rather, we see it as a useful addi?on to be able to "explore" variants more quickly, beNer and more transparently and to be able to beNer grasp not only qualita?ve design work but also quan?ta?ve urban development designs. Used sensibly, it helps to be able to quickly examine variants at an early stage. Ar?ficial intelligence is not intended to replace the planner's profession, but to beNer prepare the basis for decision-making.” For a more specific analysis of the use of Generative AI in Urban Governance see Nelson, RJ. Naing, KM. Beroche, H. 2023. Generative AI for Urban Governance.8
