Recently interviewing a number of cross – domain designers from landscape to AI products, this is the first interview guest Shadow, a designer with ten years of work experience, full stack development, cross boundary from landscape design to AI product manager.
ZILONG: Hi Shadow, would you like to introduce yourself briefly?
Shadow: I am Chi Zhiwei, Shadow, typical slash youth. 08 years graduated from landscape design major of Shanghai Traffic Design Institute, master of Tongji University. Now, the identity is a cross boundary designer, engaged in landscape design, tourism planning, real estate design management, parameterized design, user experience design, data visualization design, and so on. At the same time, I am also a full stack developer. In recent years, I have learned deep learning related keras, back-end node.js and python. Now the most famous ARKIE owner in the design circle is a product manager / machine learning researcher, and a part-time teacher in the landscape design course at Shanghai Jiao Tong University and a technical consultant for some friends. In the past few years, also in the media, the public called Mixlab, known as the column “artificial intelligence design practice guide”, has formed 500 people + designer & programmer cross community community.
ZILONG: when did you start learning by yourself? Why do you want to learn to develop? In my understanding, self-learning is not an easy task for designers.
In Shadow:08, I was learning Python and visual programming as soon as I graduated. In many landscape projects, I would adjust the 3D design in CAD and Sketchup by programming. In 13 years, I began to learn the front-end development, and then went to ZTE to serve as a senior software development engineer, mainly through Node.js to develop Hybrid APP, and to learn React Native in 2016. In the past 17 years I have changed a job. It is very big to do user experience design in China Merchants Bank, from landscape design to programming development to user experience design. I also do research and development at the same time, I want to help the bank to realize a Luban system of Ali, it can automatically generate a variety of Banner, posters, so I have learned the knowledge of deep learning. Anyway, there are new things I will try to contact and learn.
ZILONG: shame and shame. My computer background designer doesn’t have many programming skills. Why do you want to be a Luban system?
Shadow: I hope to achieve design through intelligent way. I did parametric design when I was doing landscape design in 08 years. In 2017, I spent a lot of time and energy studying the direction of intelligent design. When I left China Merchants Bank this year, the system I developed already has a available version that can see the specific effect directly, and it is not a problem to generate a Banner.
ZILONG: in the first half of 17, I saw the products of ARKIE. I think the improvement of space is quite big. Do you think you are doing better than they do in China Merchants Bank’s automatic generation design system?
Shadow: Well, it’s more subjective. It depends on which method you use. ARKIE wants to do a sentence to generate a poster. The method they used needed a very experienced designer to give different templates and rules, such as color matching, typesetting, fonts, and so on. At that time, I studied the main practice of the 2017 ARKIE, and it was the same principle. By dynamic and parameterized templates, 100 designs could be generated by a template. As long as the quality of the template provided is high enough, the effect of each Banner can be guaranteed. But Luban’s approach is different, it is based on all of Ali’s Banner data for machine learning, abstracting the corresponding rules.
ZILONG: I heard that you developed quite a lot of APP in your spare time. Can you share with us about it?
Shadow: no problem. A few years ago I did a LBS – based Postcard application called Spyfari, which was the first time I developed it with React Native, and the whole development took a long time, about three months. As long as you take a picture, it can automatically generate one sentence according to your geographical location and synthesize a postcard. How does this sentence be generated automatically? By GPS, I can determine where the user’s location is, and then display the preset material, including a variety of poetry, all of which are found by the crawler. I’ve also tried a local snapshot application that combines images from the entire site into a long one and automatically adds words to make a poster. By the way, I also do chat robot ACE Land, which is an artificial intelligence assistant App that recommends content according to user time. This APP basically calls the interface of Turing robot, but it turns out that this is not the main direction I want to do. In other spare time, I also did some small program development. I like to do some “graph + text” combination, and I pay more attention to reducing user input by automation. Users need to enter a picture or hit a few words, so that the user’s operation cost can be reduced to a minimum.
Zhirong: Actually, it didn’t take a long time for a person to develop an application. I used to develop an application at the same pace. (at that time, “three months’ development time was very long”. What do you think the design is after such a long design?
Shadow: first in a question, I think the design has two kinds of state, a kind of Party A himself, at this time you will enjoy the process of design and thinking, you can look at the problem from different angles, do not consider too many commercial problems, such a design is more pure. Another kind of Party A is someone else. At this point, I have to think about what Party A is thinking and design is relatively limited. Back to the point, what is the design? I think it is to apply some of your design “raw materials” to solve problems skillfully. The design of the “raw material” includes the skills you master, the materials used in the landscape design, the psychology of user experience design, the process of interaction, and even the code developed. It’s like buying different kinds of food in the vegetable market and making a dish through various cooking methods. In this decade I have done a variety of different designs, I think the raw materials can be different, but the method and nature are the same, the design thinking is the same.
ZILONG: I agree with you very much. I think designers should have a lot of skills. “Skill” means design thinking. “More can” be similar to the raw materials you say, extensive skills and knowledge, so that you think more fully when you do the design, and solve this problem from different aspects by designing thinking dimension. . Next question, you have ploughed so long in the field of AI&Design. What do you think is AI now?
Shadow: the problem is quite broad. How to say, AI now depends on your intelligence to what extent. It can be mentally retarded, but it is also one of AI. The so-called “very weak intelligence” means to solve this problem through very simple rules and methods, but some of the data processing I may have used deep learning, so it is also artificial intelligence, but it sounds not that high end. Now a lot of people in the industry like to say that they have used against generating networks or deep learning when solving problems. No matter what you use, the problem you solve is the same problem. It’s just a better way to assess the effect of the final evaluation. So I think what AI is, I think it is just a technical means, it is parallel with the design.
ZILONG: Well, it makes sense. I thought AI was a design method. Design is used to solve problems. Deep learning is also one of the ways to solve problems. Next, do we want to talk about AI and design?
Shadow: OK. Have you ever heard of a company called little library technology? It uses artificial intelligence to do architectural design, but the principle behind it, the way it is implemented is very different from ARKIE using artificial intelligence to generate posters. Artificial intelligence in building direction focuses more on the construction of knowledge map and the classification of CNN.
ZILONG: why is the knowledge map of architectural design necessary?
Shadow: because there are many specifications in the building. For example, a community, its height should be how much, the volume rate is what, the unit size and area of each suite, each house has a few rooms, and the area of each room is also defined according to the overall area. Behind these are strong norms and requirements.
ZILONG: so ARKIE does not do these specifications and knowledge maps, because the design includes subjective factors and abstract, it is difficult to use the specification to construct the aesthetic knowledge map.
Shadow: Yeah, I thought about doing an aesthetic map of knowledge when I was hiring, but it’s hard to classify knowledge. For example, is the word “symmetry” in terms of layout or visual balance? It is difficult for me to define the nodes of each knowledge and their relationship. But the field of architecture is not a pure design direction, it has its own specifications and requirements in many aspects. They are mandatory, so it is possible to make a knowledge map.
ZILONG: I have seen a previous case of learning to transform the chassis of a car through a machine, called Hack rod first 3D to print a car chassis, and then get different real data through a variety of sensors in the car, and then let the machine constantly learn and automatically change the chassis structure in the virtual environment. I would like to know if architectural design can be realized by similar methods and knowledge maps.
Shadow: this method is not very realistic in building design, because it is necessary to build the building first, and the cost is very high. You’re talking about a more data-driven approach, and now landscape design and architectural design have similar thinking, such as parametric design. But at this time, the model has been designed and will not be improved. If we want to achieve improvement, we need a simulator to achieve it. This is one of the difficulties. According to my understanding, structural design is a simulator, because the mechanical simulation system has been very mature, such as the design of the bridge, can be continuously simulated and constantly adjust the parameters to make the bridge design to the best state. However, there are many factors to consider in architectural design, such as how many people can accommodate, how people flow in each time period, and various subjective factors, including design sense, commercialization, and personal preferences of Party A. Architectural design is not a purely rational design, so it is difficult to combine these factors together for simulation.
Shadow: let me give you another example of jewelry design. Now, the acquisition of user data is becoming more and more simple, with the maturity of 3D printing, nanoscale and other technologies, it is more possible to customize the jewelry with artificial intelligence. The traditional jewelry design process is long. The designer needs the customer or his own creative inspiration to draw a sketch of the design. It is constantly modified by the blueprint, and then makes the jewelry template based on the jewelry design, then the handmade wax version or the computer CAD, then the mold, the die, the mosaic, the polishing and the surface treatment. Final quality inspection and certificate. Artificial intelligence jewelry designer can do a lot of things before giving the final jewelry design, such as letting the machine get the user’s personal data, including sound, height, weight, heart rate and personal preferences, as well as the material, symbols, weight and other jewelry parameters selected by the user, and then visualize the data and combine them with each other. Pass the algorithm to generate different design, and finally let users find the favorite 3D jewelry model. The artificial intelligence jewelry designer can even make the user shape the jewelry by itself. The user only needs to draw a rough model and use RNN to display the 3D jewelry model which is the most matched user.
ZILONG: I see. Let’s change the subject. Sometimes I really feel that I don’t know how to design technology, so the design will be very limited. It’s hard for you to see clearly what the essence of your product is and what the frame is. What do you think is the ability of programming development to help your design?
Shadow: ha ha, the simple point is that understanding development can make your design more technical. Let me give an example of a chat robot. If you don’t know how to develop, you don’t know how to evaluate the effect of a chat robot, and you don’t know what to do to improve the effect. If you are an ordinary designer, you may think that all the chat robots are so high, so good to use online, and then the products of your home are designed as intelligent, but in fact, there is no meaning, because you can’t do it. But the average designer would think that this could be done, because this kind of gossip is understandable to humans, and other competitors do the same. When your product concept is divorced from the practical implementation plan, you will never achieve your product goals. Give another example of a filter. If it’s a designer, he might feel like using Photoshop to add a very cool filter to the picture, and then give it to the programmer to make them come true.
ZILONG: the case of the filter is quite right. I’ve been doing camera related products in the company before, and basically all of the filters are implemented in open source code. It’s not realistic to rewrite a filter because few engineers understand image processing. Although the performance of the filter is very related to the designer, it doesn’t have much to do with the designer, because the Engineer you consider is probably not able to do it.
Shadow: let’s talk down the filter. I’ve been looking at the implementation of the filter recently, and the better filter effects are written by the GPU shader. If it is conventional image processing, the efficiency of using pixel processing method to filter will be too low, and the style is small. But it’s really hard for many engineers to write filters with a GPU shader, and not all engineers know how to develop shaders. And the shader function is very powerful, how cool it can be even designers do not know.
ZILONG: Well, yes. I’ve written front – end related code, and I believe a lot of engineers can write page code, but great dynamic code is not written by all front-end engineers, because they don’t learn this knowledge. Moreover, a particularly dynamic effect is more of a combination of design and development, which is interdisciplinary. There are many engineers who have not learned SVG, SVG I have just seen some, it is just a file format, in fact, can do a lot of things, including a variety of complex animation. When I wrote my own website two years ago, I also used SVG animation to do it. It was really complicated. I can only watch the source code of others slowly change to the effect I want, but it is unrealistic for me to learn and develop SVG from 0 to 1, because I really don’t have time.
Shadow: Yes, this involves knowing whether you want to focus on a specific field or all areas.
Zhirong: The appearance of Lu Ban last year has caused many designers on the Internet to worry about being eliminated. What do you think of the relationship between AI and designers?
Shadow: I think there are several main relationships between AI and designers. One is a purely labor-intensive designer who only knows how to copy, paste and alter drawings, and this designer is likely to be replaced. There is also a designer who specializes in his own field. Such designer AI may not have much to do with him.
ZILONG: I don’t agree with you. It’s like copying a picture, some people have spent a long time copying it, and I think it’s more like a deep ploughing rather than a pure labor, but AI may have a very good copy of the copy in a style.
Shadow: Well, but this is more art, art is not an industrial product, industrial products will be efficient, the art you want is to make machine production or people to create, which is worth thinking. I have other ideas recently. A designer in a certain field can quickly build its own brand in this field. He has a congenital advantage. Even if AI is strong, it is hard to compete with him.
ZILONG: that’s right. I think the techniques may reach the ceiling for the designers, but your idea and influence are the most important. I have a feeling here. When I started to learn to interact in 12 years, I had read all the books on the market 14 years ago, but after 15 years I found it difficult to find new interactive books, because the interaction design was almost the same, so the books were less. When each person’s interactive design skills are almost the same, it is more important to think about how to improve their other abilities, such as understanding of the business, how to expand the ideas and techniques of other areas.
Shadow: Yes, so it’s hard to define the relationship between AI and the designer. It depends on how the designer is located, whether he is transboundary or only a little. AI is just a tool for cross – boundary designers. But this kind of cross-border talent has been difficult to define with the designer profession, I think he is a higher level than the designer.
ZILONG: Yes, let’s talk about the last question. How do you think designers should broaden their horizons?
Shadow: the most important thing is mentality, mentality must be open. No matter which field or content, you must contact them with an open mind, contact you and give feedback later. You can’t be particularly disgusted with the opinions put forward by others or the experience accumulated in other fields. You don’t think you’re right. If you want to think so, the other person may be right. I want to listen to it first and then consider it comprehensively. Considering every point equally, I think it’s easy to broaden my horizons and abilities, but it’s hard to do that. It is also to communicate with people in other industries and to communicate with people of different experience, such as young college students or elders of fifty or sixty years old, to expand their horizons when chatting. The Mixlab community I created is also for this purpose, so that people from different industries can learn from each other, learn from each other and make progress together.
ZILONG: Thank you very much for Shadow today’s sharing. I hope to have the opportunity to come to Shanghai to participate in your Mixlab gathering.