计算机科学与风景园林——弗吉尼亚大学风景园林系系主任耿百利教授专访

采访：张诗阳

耿百利教授是弗吉尼亚大学风景园林系系主任，曾任教于哈佛大学、路易斯安那州立大学、罗德岛设计学院。他关注于计算机技术、人工智能在风景园林不同领域中的应用，并开展了大量相关的研究、教学与实践。他曾受邀参与在北京林业大学举行的2019年世界风景园林师高峰论坛，《风景园林》杂志社有幸对耿百利教授进行了专访。访谈中耿百利教授结合自己的大量研究，探讨了计算机科学对于理解环境与生态系统、风景园林设计及教育的影响。以下是采访全文。

LAJ：《风景园林》杂志社

Bradley：耿百利·坎特雷尔

LAJ：您的本科专业是计算机科学，而后在哈佛学习了风景园林。是什么原因让您选择了风景园林作为职业呢？

Bradley：这是一个非常好的问题。在北美很多人并不熟悉风景园林，直到他们以某种方式接触它。在学习风景园林之前，我学习的是工作室艺术（studio-art）和计算机科学。我了解风景园林纯属偶然，起初我并不熟悉它，我热爱植物、热爱户外，同时也喜欢电脑制图。因此，当我发现风景园林可以将以上方面融合到一起时，我发现它完美地适合我。事实上，我是在一家植物种植公司了解到风景园林的，有人告诉我你可以用植物进行设计，所以我开始接触风景园林。

LAJ：我相信您的计算机背景对您近年来的风景园林研究有很大的影响。您更多地借助智能生态系统、数字模型等来实现自然的自主性，这与这些技术在建筑设计或城市规划领域的应用截然不同。您认为数字技术与风景园林融合的关键词和特征是什么呢？

Bradley：我认为数字技术在建筑、规划、甚至是工程领域的应用多数时候是在寻找一个特定的答案，这与风景园林领域有着很大的不同。通常人们需要通过计算明确地了解到对象的本质，从而找到最有效的方式来解决问题。然而，在风景园林以及环境设计领域，我们通常也运用计算机去了解系统的功能，但是人类与环境的相互作用是十分复杂的，包括生物的复杂性、其他物种以及所有类型的环境过程和功能，这些都不太容易进行运算。因而我们希望开发一种新的计算机应用方式，其不仅能够理解自然和环境过程的运转模式，也能直接进行互动。所以，我认为最大的不同在于我们必须开展一对一的工作，逐一构建模型并开展与环境的互动实验。这些实验使我们能够得到环境的反馈信息，然后借助计算机来理解这些信息并开展行动。但这并不意味着我们必须了解全部，只是我们开始运用计算机的力量去与环境进行互动来进行学习。

LAJ：我注意到您的很多研究谈到了人工智能，您认为当前人工智能在风景园林行业应用的关键是什么？人工智能将在处理环境问题中扮演怎样的角色呢？

Bradley：这是一个宏大的问题，可能会讲述好几个小时。目前我们对环境的理解通常取决于2个方面：一方面是环境的自身作用，比如树木或者物种的自然生长，我们称之为自然，或者野性；另一方面，我们自身也维护、建造或者设计环境，我们通常称其为景观，也就是说通过人类的介入来改造环境。我认为目前人工智能可以帮助我们思考如何来维持景观和环境，将会用有一种智能维护的景观，我们将创建一种新的环境类型，它具有高度的维持性，我们也许永远不会真正地了解，但是把它看作是一种自然或者 野性。

因此，环境维持的方式以及我们与计算机智能的关系将发生根本的变化。有一件事情目前还没有理解清楚，在我们利用机器学习了解景观新事物的同时，我们目前还不能确切地了解它是如何完成它的过程的。我和张子豪（弗吉尼亚大学在读博士研究生）都写过相关的论文[1-2]，即第三类人工智能。我们属于生物智能的一部分，但人类总喜欢把自己刨除在外，从而出现了人类智能，当然还有计算机智能（人工智能属于其中的一部分）。这其中包含3个角色，我认为它们每一个都在我们理解如何将人工智能运用在社会或者风景园林时扮演着不同的角色。

LAJ：所以你说第3类智能不再是关于自然的？

Bradley：它已经并且将会从根本上改变人们思考自然的方式。有的地方我们已经通过复杂的程序来维护它们，它们并非没有受到影响，它们在持续发展，实际上有很多过程在变化。维持这种变化的智能不是人类智能，我们并不真正理解为什么电脑会这样做，因为计算方面的东西对我们来说是未知的。因此这创造了一种新的关系和一个新的生态系统。

LAJ：从您的观点来看，人工智能应该作为风景园林的媒介来构建连接人、自然和技术的框架系统。您能给我们介绍一下在这方面的实验或实践吗?

Bradley：考虑这个问题的一种方式是把许多因素结合在一起，因为人类把所有这些自然和技术，以及不断发展的文化都联系在一起。我们一直在寻找这样做的方法，这是设计的核心。当人们做出这样的努力时，我们会非常认同，因此通常它可能是生态的，但也可能不如美学重要。有一种奇怪的关系，设计师总是倾向于人性而非与我们关注的一致。设计会有很多其他目标，那么结果可能是我们开始为其他物种构建其他类型的景观，其中人类是次要的。它们有高度维持性，但是是为了服务其他物种。

LAJ：您认为人工智能的干预会对当前的风景园林设计过程产生怎样的影响?

Bradley：我想我可能说了很多，简短地说，我一直是在一个长期的视野里探讨，也许20年以后。但在短期内，我认为我们所看到的最大变化是以一种新的方式来思考景观形式和景观模式，这是我使用机器学习的主要方式。它是一种非常复杂的系统，机器学习可以在这些模式中找到相似之处，并开始思考形式的内涵及其与其他系统的关系。美学的形式也许不同，但它们运转方式都是相同的。机器学习可以帮助我们透过形式看到其中的内涵。我认为在短期内，它将改变我们的视角，从形式的研究转向为现象的研究。

LAJ：所以您的意思是我们应该决定我们应该设计什么？

Bradley：为它设计一种新技术，对吧？它不只是我们看到的形式。

LAJ：它不仅是一种形式，更是一种策略或者其他方面。设计是一个复杂的过程，既包括科学的逻辑思维，也包括对场地的感性认知。您认为数字技术在环境感知或场所艺术性方面扮演什么角色?

Bradley：实际上我并不认为数字技术会带来太大的改变。在某种程度上，我们仍然是人类，我们仍然在用我们的工具以类似的方式进行设计，主要的区别是2个部分：一方面，数字设计的一个优势是速度，我们可以快速地进行筛选和生产；另一方面，我们可以进行更精确的分析。其缺点可能是会为我们带来太多的选择，并削弱实用性。所以我们必须敏锐地意识到，我们仍然需要在场地中建立一些切实的联系。数字化使得理解场地的物理属性变得更加复杂，因而我们需要开发一种新的方式来精确地进行场地介入。

LAJ：最后，让我们回到教育。风景园林教育中传统的计算机技术教学更多地涉及软件学习，如CAD、Rhino等。在计算机相关技术快速发展和人工智能逐渐成熟的背景下，您认为在当前的风景园林教育中应该如何开展计算机相关的教学?

Bradley：我的观点是，媒体教学需要有它自己的位置，它不能只服务于设计工作室。类似于风景园林历史，我认为数字技术在场地建造、生态、媒体或者其他方面的应用都应该在课程体系中有它的位置。我知道很多项目都与其工作室联系在一起，但我认为目前已经能够确定我们不仅需要了解如何使用这些工具，同时也要学习这些工具是如何产生的以及它们的历史。关于教学工具，目前有很多资源，我们可以在Youtube上进行学习。我认为有一点十分重要，就是我们不仅应该教授学生如何使用这些工具，更应该教给他们一些基本方法来更加有效地利用它们。在每一项训练中，我们都要求学生运用一个小的工具集对景观进行概念化的定位。这是通过景观本身的方式来理解工具的使用方式。不仅仅是如何表现，更是探讨它们对于土壤意味着什么？树木是如何生长的？所有这些方面的教学内涵说明我们不仅应该教授如何绘制图案，而且要借助这些工具来探索事物的本质。我认为把这2件事结合起来是非常困难的，但是也是很重要的。

LAJ：因此，媒体应该帮助学生理解景观，理解自然是如何工作的，这都非常重要。

图片来源：

图1由耿百利、Jeff Carney、Matthew Seibert绘制；图2由耿百利、Cheramie、Seibert、Carney绘制；图3由耿百利拍摄；图4由哈佛大学设计研究生院提供。

录音整理：赵文迪

（编辑/王亚莺）

LAJ: We know that you majored in computer science as an undergraduate, and then studied landscape architecture at Harvard. Why do you choose landscape architecture?

Bradley:That is a good question. Well, particularly in North America, not many people know about the landscape architecture, until they come across the discipline some how. I studied studio-art and computer science, and then majored in landscape architecture, when I found out about landscape architecture, it was by accident, I didn’t know about it, I like plants, I like the outdoors, but I also like computer graphics. So when I found landscape architecture and all those things come together, it was perfect for me. I found it because I was actually working for a company planting trees and someone told me you can design things with plants, so that is how I found it.

LAJ: I believe that your computer major background has an impact on your researches over the years. You have applied robotic ecosystems or digital modeling more to achieve the autonomy of nature, which is quite different from applying these technology in architectural design and urban planning. So are there any key words or features in your mind regarding the integration of digital technology and landscape architecture?

Bradley:Well, I would like to say there are some key differences between how we deploy technologies in architecture, planning, and even when we look at engineering, most of the disciplines are typically looking for a specific answer. You need to calculate it, you know exactly what it is and then you try to find the most efficient way to come up with that answer, often times, in landscape architecture, and even in environmental design, not only are we using the computation as a way to understand this system’s function but we also have the complexity of human interaction with the environment, and then we have the complexity of biology, other species, and all types of environmental processes or functions, and it is not so easily computable. So, what we see is that we have to develop a new way of using the computer, not only making the models that allow to understand the way nature and the environmental process works, but also allows it to interact directly. So what I believe the biggest differences is that there has to be a way that we work one to one, with the way we create the model, and the way we interact with the environment itself. When we do something to the environment, we get the feedback from that, we allow computing to help us to understand that, and we are able to act on that information, and that doesn’t mean we have to understand everything, but we begin to learn through the power of computation and the interaction of the environment.

LAJ: I have noted that many of your papers talk about artificial intelligence. What do you think is the core challenge of applying artificial intelligence in the current landscape architecture profession? Or what role will artificial intelligence play in dealing with environmental issues?

Bradley:well, there are two parts of these. I mean, well, this is a big question. I think, I could probably talk about this for a few hours. But we just break it up two parts. One is that, our current understanding of the environment essentially depends on two aspects. One, the environment acts in one way, like trees grow certainly and species acts certainly, and we call that nature or even wildness. On the other side of that, we also maintain, construct, create the environment, and often times, we call that landscape, that is, we as human beings intervene to construct it. I think the thing we will see from the forms of AI, particularly, when we are thinking about how to maintain landscapes and the environment, we are going to have another intelligence maintaining landscape and creating a new type of environment that we really never understand, it is highly maintained, but we view it as natural or wild.

So, there is a fundamental change in the way environments are maintained and our relationship with computational intelligence. One thing that we do not quite understand yet, is that not only can we use machine learning to tell us new things about the landscape, but we are also getting to a point where we don’t understand exactly what or how it is accomplishing its procedures. Zihao Zhang and myself have written papers on this idea, that there is actually a third intelligence[1-2]. We have biological intelligence, we are kind of included in that but human beings like to separate themselves, and there is human intelligence, and there is a new computation intelligence, and AI is part of that. Within that there are also three actors, I think, each plays a different role in how we begin to understand, the function of AI on society, as well as in landscape architecture.

LAJ: So you say there is no more about nature, as the third intelligence?

Bradley:It has and will fundamentally change the way you even think about the nature, because we have places and we are already maintaining through complex procedures and they are not left untouched, they are persisting, there are actually lots of procedures intervening there. The intelligence that maintains those kind of changes are not a human intelligence. With that we don’t really understand why a computer does this, because the computational aspects are hidden from us, therefore this creates a new relationship and a novel ecosystem.

LAJ: I have found that from your perspective, you think artificial intelligence should act as a medium in landscape architecture, which is to build a structural system linking people, nature and technology. Could you brief us some of your experiments or practices in the context of this concept?

Bradley:One of way to think about this is in terms of pulling together many elements, as human beings are linking together all these nature and technology, as well as an evolving culture there. We are always trying to find ways of doing that, it is at the heart of what design does. When people take on this effort we favor human beings, so typically, it may perform ecologically but that might be secondary to aesthetics. There is a kind of strange relationship where designers are always favoring humanity, rather than favoring us all the time, it might have other goals, so the result may be we start to have other types of landscapes that are constructed but perform for other species and humans are secondary, once again, highly maintained but productive to other species.

LAJ: What effects do you think artificial intelligence intervention will have on the current process of landscape architecture design?

Bradley:I think I probably said it a lot, so I’m going to say it in a short term. What I’ve been talking about is in a long term horizon, maybe two decades away. But in a short term, I think the biggest change we are seeing is a new way to think about the landscape form and landscape pattern, and that is the most of the way I’ve been using the machine learning and. It is a way of taking very complex system and allowing machine learning to find similarities in those patterns, and start to think about what is in the form and its relationship to other systems, the forms aesthetically may be different, but they all perform the same, and the machine learning helps us with that, because it is not just the form that we see. But it is about what the system performance o, so I think in the short term it will change our perspective from formal inquiry to the performative inquiry.

LAJ: So you mean we should decide what we should design?

Bradley:And design a new technology for it, right? It is not just a form what we see.

LAJ: It is not just a form, it is a strategy or some thing else. The design is a complicated process, which not only includes the scientific logical thinking, but also the perceptual cognition about the site. What do you think is the role of digital technology in the perception of the environment or the artistry of the site?

Bradley:I actually don’t think digital technology changes that much for us, I think in some sense, we are still human beings, we are still using our tools to perform a design in similar way, the major differences are two parts. One is an advantage of digital design, is that of speed, we can make alternatives, we can have choice, much quicker, we can produce very quickly, and the other advantages is that we can develop an analysis that are more precise, and the disadvantage is that we have too much choice, and tangibility starts to disappear, so I think the change is that we have to be acutely aware that we still need some tangible relationship of the site, so the digital makes it more difficult to understand the physical place, so to develop a new way of doing precise site engagement.

LAJ: At last, let’s return to education. Traditional computer technology teaching in landscape architecture education involves more with software learning, such as CAD, Rhino and others. Under the context of rapid development of computer-related technologies and gradual maturity of artificial intelligence, how do you think the computer technology related teaching should be carried out in the current landscape architecture education?

Bradley:So my take is that the teaching of media, it needs it’s own place, it cannot just serve the design studio, so I think similar to the landscape architecture history, in terms of teaching, site construction or ecology, media or other kinds of the computation. They all need their own place in the curriculum. I know many programs tie directly to studio. but I think there is enough discourse now that we not only need to know how to use these tools, but also know where do these tools come from, what is their history. The other part is that teaching of the tools, there are so many resource out there, I mean, we can learn these tools on the Youtube. But I think there is a one thing really important is that we are not only help students to learn these tools, we give them some fundamental ways to use the tools in an effective way. We also immediately ask them to take the small set of these tools and conceptualize the role of landscape in every exercise we give to them. This is a way to understand the tools through the media of landscape. And not only how we represent it, but what it means to soil performance, how trees grow, all of these aspects need to be inherent in the teaching of software, it is not just about making images, it is also about how to use the software to explore the actual physical landscape itself. I think layering these two things together is a really difficult task, but I think layering these two things together is really crucial.

LAJ: So the media should help the student to understand the landscape how does the nature works is very important.

Sources of Figures:

Fig. 1 © Bradley Cantrell, Jeff Carney, Matthew Seibert; Fig. 2 © Cheramie, Seibert, Cantrell, Carney; Fig. 3 © Bradley; Fig. 4 © Harvard GSD.

Recording collector: ZHAO Wendi