AiCorb: Tackling the Designer's Dilemma and Risks with AI-Powered Rapid Prototyping

YOSHITO TSUJI: Today, we would like to introduce AiCorb, our new AI design assistant developed by Obayashi Corporation. Also, we will talk about the concept of AI-powered rapid prototyping, a new design process using AiCorb. Hope you enjoy our presentation.

Let me briefly introduce Obayashi Corporation. Obayashi Corporation is Japan's leading general contractor, operating in Japan, United States, and APAC regions over 130 years. In addition to construction, Obayashi has expanded our business field into urban development and renewable energy.

What is AiCorb? AiCorb is an AI Design Assistant. However, not just a design generator. Gaudi said, "Human do not create, they discover, and upon those discoveries, they build." We think AiCorb will be a powerful assistant for exploring discoveries for architects.

This is a key takeaway of our presentation. Our new proposed design process, AI-powered Rapid Prototyping, will enable early client agreements and increase their satisfaction. Also, it will solve design dilemma and reduce project risks. Also, it will bring broader stakeholder info-sharing and social benefit. By AI-powered Rapid Prototyping, we realize better informed decision-making with all stakeholders.

Before we jump in more detail, let me introduce us. Myself, Tsuji is an architect and working at Obayashi Corporation for 30 years. Designed R&D center, office building, and high end boutique and so on. At the same time, I also Chief Innovation Officer of Obayashi Construction Tech Lab Singapore, which is promoting robotics construction in Singapore and other APAC countries. Let me introduce Takuma.

TAKUMA NAKABAYASHI: Thank you, Yoshito. My name is Takuma Nakabayashi, and I'm an AI researcher responsible for AI-related technology development at Obayashi Corporation. I spent about three years here in Silicon Valley, California, conducting research and development for the AiCorb project that I'll be introducing today. It's a pleasure to be back here and to have the opportunity to present AiCorb to you. Now, I pass back to Yoshito.

YOSHITO TSUJI: OK. Thank you, Takuma. Let us start by telling you why we developed AiCorb. I will be the client, and Takuma will be the architect, so you can feel familiar more with our story. Please enjoy it.

[KEYBOARD CLICKS]

[CLICKS]

[EXPLOSION]

Enjoy. Have you ever had an experiences like this? This is everyday life for architects. Even I bought my idea that I was very confident. But sometimes they are completely defeated by clients. Very difficult to convince clients.

In the early design stage, usually it is not clear what client want to build. Even in such a situation, architect should explore the client intention, sometimes hidden idea by presenting several different options, considering building codes, costs, and other factors.

For medium sized office building case, these composition during the feasibility study often takes a couple months, but sometimes all the steps can be wasted. How quickly you agree with a client is very important.

If we draw this situation in the diagram, it would be like this. Although the architect is always creating design options, client decision-making is always delayed. So usually, there's a large gap between creation by architect and decision-making by client. This gap would be enlarged in concept design stage and early basic design stage. This gap became design dilemma and project risks.

In my previous project for US IT client seven years ago, I always require to present several options, and I consider every possible option. At the end of the project, my hair came completely gray, like this. Not like this manga guy.

Dilemma is categorized into two. One is dilemma of creativity. The other is dilemma of efficiency. However, each dilemma appears differently, depending on whether you are looking at it from client side or architect side. We always have give and take between the two sides. Let's look at more detail.

Architect explore unlimited option within limited time, and client usually ask architect to propose something new and interesting. Many architect refer to past and similar examples. In my case, I have sometimes copy and paste my previous work and sketch over it.

Using my hands, allow me to experience that joy of creation. Physical cut and paste is faster and better for me, but other persons prefer other way. However, in general, there is not enough time to explore new and interesting ideas, and architect must keep thinking. We must do quick and quick.

Dilemma in efficiency is even more complicated. Even if multiple options are selected by client, to obtain specific figures, such as costs, schedule, and environmental score of each option, is so tough. If architect push design team so hard, often everyone say, cannot be done until other profession finish.

Sometimes argument aligns between architects and engineers. In many cases, most of figures are not available until the end of the design stage, so client cannot get sufficient information for their decision-making.

We developed the concept of AI-powered Rapid Prototyping to fill the gap between creation of architect and decision-making by client, by solving these two dilemma and reducing project risks by AI.

Focusing on concept design stage. Our AiCorb will be a design assistant that enables architect to provide clients with the right information at the right time. By this, client can get a better informed decision-making. OK, let me pass to Takuma to explain about AiCorb capability. Takuma, please.

TAKUMA NAKABAYASHI: Thank you. So, from now on, I'm going to explain the overall structure of AiCorb to resolve the dilemma and risks we've discussed so far. We propose a method called AI-powered Rapid Prototyping through the development of AiCorb. But first, please watch this concept video about AiCorb.

[VIDEO PLAYBACK]

- Creation continues uninterrupted through humans. But humans do not create. They discover. And upon those discoveries, they build. Antoni Gaudi. In pursuit of a better future, we give design to our discoveries to build an ever new world. Those who came before us were in constant progress.

How did they turn their dreams into reality? As architects, dialoguing with our clients, we strive to make the best possible design proposals through a process of trial and error. A conversation between design and oneself has provided a key to solving the profound dilemma between creativity and efficiency. Now, in the age of AI, we're developing an innovative new design process, the AiCorb.

The AI in AiCorb instantly generates a variety of design ideas from a simple sketch. It can read the intentions behind any sketch and reflect those in a design. The generated image can then be modified directly by applying further sketch or design instructions, leading quickly to an ideal design.

Additionally, the AI within AiCorb can generate a 3D model from that ideal design. More easy to conceptualize than a simple picture, a 3D view facilitates discussion of various factors, like environmental or cost analysis.

AiCorb delivers appropriate information at the right time, empowering clients to make the best decisions. This not only solves the architect's dilemma, but also promotes better informed decision-making between clients and other shareholders.

It is a revolutionary solution that assists architects in the balance between creativity and efficiency. The process of design approaches new horizons with AiCorb.

[END PLAYBACK]

TAKUMA NAKABAYASHI: OK, thank you for watching the video. Since its past press release in 2022, AiCorb has been featured in numerous media outlets. It was included as a leading example of image-generating AI in Japan's Information and Communication white paper, and was even used on a TV variety show to design the ideal homes for comedians.

The AiCorb 3D modeling function that I'll introduce shortly, is expected to be available on the Autodesk App Store in early October. I'll provide more details on this feature later, but with this plugin, you can quickly convert generated proposed design concepts into 3D models. Please give it a try.

Before introducing the specific feature of AiCorb, let me explain how we use AI to tackle the dilemmas and risks we mentioned. This is a traditional architectural design workflow. First, based on site conditions, regulations, and the client requests, we consider the building's volume, floor plans, and elevations.

Once these are decided, we will move on to the interior and exterior designs. Design isn't just about appearance. It must also take into account factors like cost and the environmental performance. To evaluate these, architects typically create 3D models manually and run various assessments.

Based on the results, they then adjust the design and make proposals to the client. In the process of building consensus, this cycle is repeated again and again. And as just mentioned, many proposals end up being rejected in this cycle.

The issue is, that design and 3D modeling will take considerable time, and having to repeat this process again and again. Let's do the limits and risks. To help reach a satisfying goal with fewer iterations in the consensus building cycle, we developed AiCorb.

Currently, AiCorb serves as a tool to assist in the design and model stages. With AiCorb Facade Design Function, architects can generate various design options from sketches using generative AI based on the architect's intent and the client's needs.

The AiCorb 3D modeling function helps turn these generated images into 3D models. By using these AI features, architects can present new proposals to clients much earlier in the process.

Now, let me explain more concretely how AiCorb has been addressing these dilemmas and risks. First, please watch this video on the Facade Design Function.

[VIDEO PLAYBACK]

- Design decisions require ongoing dialogue between the client and architect. With AiCorb Facade Design Function, you can visualize the client's ideal design from the first meeting. AiCorb Facade Design Function is intuitive to use. Even if you are not an architect, it brings your design ideas to life.

You can explore further from your favorite design. You can add note to the sketch and make partial edits to the generated design. Design exploration used to start with searching real world examples, but now AI lets you generate entirely new ideas. AiCorb helps you make design decisions with greater satisfaction.

[END PLAYBACK]

TAKUMA NAKABAYASHI: OK, thank you for watching the video again. As shown in the previous video, AiCorb places greater emphasis on sketches. Our AI is specifically trained to understand an architect's intent, even from rough sketches.

Why do we focus so much on rough sketches? The result, the dilemmas, and risks faced in the early stages of architectural design. We believe this function must be something that can be used quickly during meetings.

With the rise of computational design, some architects prefer exploring possible designs using 3D models. However, making quick adjustments during a meeting is challenging. Adjusting a few parameters might be possible, but changing the overall feel of a design is not easy.

On the other hand, sketches have no such limitations. They allow for quick changes to the building's volume, or adding symbolic details to the facade. And that's why we still believe that sketches are the most effective tool for architects to quickly express their intent during discussions, especially when design changes are needed repeatedly.

With AiCorb, various options can be generated, even from rough sketches. We believe that the sketch to design flow using generative AI is the optimal method for optioneering, which is crucial in resolving dilemmas. It's rare for clients to fully articulate their design requests at the start of the project, so it's common for architects to present numerous options to them.

However, preparing options that aren't selected also takes time. Now, this is a real dilemma architects face. They often end up creating more design options than are really necessary. But if you can confirm the client's real preferences during the past meeting, this dilemma can be resolved.

The benefits of using generative AI go far beyond just speeding up design exploration. Traditionally in the early stages of architectural design, architects would expand their ideas by referencing existing examples. But the options you can search for are always limited. And, of course, you can search for something doesn't exist.

With the arrival of generative AI, the ideation process has transformed dramatically. Now architects can generate limitless new designs without being restricted by existing examples. This means you are not just searching for ideas, but also, you are bringing the images in your mind to life, and verifying them on the spot.

Generative AI is particularly powerful in combining different elements, like mimicking specific branches in architecture or blending architecture with a particular culture style, era, or even living creatures. This allows architecture architects to instantly explore entirely new concepts. It's a tool that will spark new inspiration in the world of architecture.

So far, I have explained how AiCorb 's Facade Design Function helps overcome dilemmas. As you've seen, generative AI can produce incredible visuals. However, we cannot propose buildings that are unbuildable or too expensive or lack environmental performance.

These are not things that can be judged from one image alone. And clients cannot make decisions based solely on a beautiful picture. In other words, a great picture alone is nothing more than pie in the sky.

So, we looked for a way to go beyond just creating images. We thought about how to use this amazing technology to better support consensus-building, and provide a practical solution to the real dilemmas and risk faced by architects and clients. This led to the development of the AiCorb Modeling Function. First, please watch this video.

[VIDEO PLAYBACK]

- For design decisions, clients need quantified information, like environmental performance and cost. They can't be understood by appearance alone. These aspects can only be evaluated once the design is turned into a 3D model.

AiCorb Modeling Function assists in converting designs into 3D models. It reads details, such as window sizes and materials, from a design image, and reflects them as Revit family parameters.

Let's give it a try. Apply this design to the lower levels. And then apply this design to the upper levels. The extracted designs are stored in the palette, allowing for easy comparison of different option.

By using the AiCorb Modeling Function, you can evaluate design proposals with numerical data, helping clients make quicker decisions. Let's start using AI, not just to create beautiful images, but to take design further.

[END PLAYBACK]

TAKUMA NAKABAYASHI: OK, thank you again for watching the video. So, as shown in the video, this feature uses AI to extract the facade characteristics and converts them into parameters, like tunnel aspect, window aspect. window depth, volume thickness, and panel color, as illustrated here.

As I mentioned earlier, this feature is scheduled to be available on the Autodesk App Store in early October, and will function as a plugin for Revit. We concluded that converting the generated images into parametric 3D models is necessary to address the dilemmas and risks we have discussed so far.

In recent years, AI technology for estimating realistic surface models from photos has become mainstream. While this is an impressive technology, and actually, I'm personally excited about it as a researcher, but we see it as an extension of image-based techniques. In other words, it's mainly focused on visualization.

What clients need for decision-making, quantitative details, such as environmental performance, cost, and various other performance metrics. To analyze these aspects, parametric 3D model is required. We adapted this approach to perform quantitative analysis and present the results to clients early in the process, ultimately speeding up decision-making for better informed decisions.

As shown here, the 3D modeling function allows us to easily convert generated parts of the images into 3D models. On these 3D models, we can extract real scalar data, such as the area of opening and the shape of the exterior. This allows us to provide approximate, but quantitative evaluations, such as sunlight exposure and environmental performance early in the process, which can be presented to the clients.

While inside, we began through our efforts and that resolving the dilemmas and risks faced by clients and architects requires an approach that optimizing both the emotional and theoretical aspects in architectural design. Although failing only one of these aspects is not enough, you can't make a decision based on just one side.

Up until now, most of the technological developments aimed at improving design efficiency have focused on the theoretical aspect, particularly those related to analysis and plan part. There are several reasons for this, but one of the main reasons, is that we simply didn't have the technology to effectively address the emotional aspect.

However, now generative AI is beginning to change the situation. Many of you have likely been amazed or moved by images, music, or videos generated by AI. But even so, generative AI by itself cannot fully bridge the gap between the emotional and theoretical aspects.

Understanding this challenge, we created AiCorb as a tool to support the proposal of emotionally impactful designs, and linked them to quantitative analysis. With this approach, we can resolve the kinds of dilemmas and risks shown here, and achieve smoother, more convincing decision-making. OK, now it's time to wrap up, so let me hand it back to Yoshito again.

YOSHITO TSUJI: OK. Takuma, thank you for a good presentation. Let me recap the key takeaway. We propose a new design process, AI-powered Rapid Prototyping. It will enable early client agreement and increase their satisfaction. Also, it will solve a design dilemma and reduce project risks. And also, it will bring broader stakeholder info-sharing and social benefit.

By AI-powered Rapid Prototyping, we realize better informed decision-making for all stakeholders. We introduce AiCorb, a design assistant, to realize this. AiCorb works on concept design stage, and enable architect to provide client with the right information at the right time.

Although development is still in progress, we are striving to realize a new design process. Our vision is that AiCorb will solve design dilemma and reduce project risks, not only for the client and architect, but also for the multidisciplinary team, by sharing information generated through AiCorb.

Furthermore, we believe that better informed consent can be created by sharing various information from the early stage of the project, not only with the project team, but also with various outside stakeholders.

AiCorb is a powerful assistant for exploring this vision. We collaborated with all these people until now. We would like to express our gratitude to SRI International, Autodesk Research, Hyper, Geo, and many individuals. Thank you so much.

As a final, me introduce our LinkedIn page, as shown here, together with Obayashi Construction Tech Lab Singapore. Please check it. This is the end of our presentation. Thank you for listening. Thank you.