Can AI help you with Building Codes?

We are excited to announce the BETA launch of our upcoming AI-powered Building Codes Copilot. As part of this program, we are releasing an initial feature called Question & Answer. You can type in any query related to building codes and get a relevant answer from our database of over 60,000 code sections. For example, you can ask "What is the minimum ceiling height for habitable rooms?" or "How many exits are required for a restaurant?" and get a clear and concise answer.

To access this feature, simply select "Question & Answer" as your search type and join the waitlist for exclusive access. The Building Codes Copilot will revolutionize the way you navigate and learn about building codes. 

Why use the Building Codes Copilot?

Building codes are complex and often confusing. They vary by jurisdiction, year, and project type. They also change frequently and have many exceptions and interpretations. Finding the right code section for your design problem can be time-consuming and frustrating.

The Building Codes Copilot is designed to make your life easier. It uses natural language processing and machine learning to understand your query and provide you with the most relevant answer from our comprehensive database of building codes. You can also see the source code section and its context for further reference.

The Trax Building Codes Copilot is not only a search tool, but also a learning tool. It helps you discover new code sections that you may not be aware of, and learn how they apply to your project.

How to join the BETA program?

The Building Codes Copilot is currently in BETA testing and we are looking for feedback from our users. If you are interested in joining the BETA program, please follow these steps:

1. Select "Question & Answer" as your search type on our website.

2. Click the join the waitlist button.

3. Wait for an email notifying you of access to the feature.

4. Start asking questions and get answers from the Building Codes Copilot.

5. Share your feedback with us through the feedback button on the website.

We appreciate your interest and support in our BETA program. We hope you enjoy using the Building Codes Copilot and find it useful for your design and construction projects. Can AI help you with building codes? Please let us know!

Challenges with Open Questions and AI Answers

There are still big challenges in trying to automatically answer open questions. Our Building Code Copilot is powered by generative AI, so surprises and mistakes are possible. Make sure to check the facts, and share feedback with us so we can improve and strengthen the safety configurations of the system! As large language models mature, the outputs will be safer and more aligned with the users' needs and expectations. For example, a Q&A system should avoid generating answers that are inaccurate, incomplete, outdated, misleading, or contradictory to the official sources of building codes.

To achieve safety and alignment of generative AI for Q&A on building codes, we will adopt industry best practices and track the latest research advancements.

Recommended Sections

The Trax Codes AI journey actually started back in January when we announced the launch of Semantic Context features that show related articles in other documents, this feature also uses the underlying AI technology that powers our Question and Answer tool to identify articles and sections in other documents that are semantically similar to the currently selected article.

With this Semantic Context feature, you can easily discover related content and improve your understanding of the context and intent of an article you are looking at. Simply select an article of your document, and click the graph button in the article action menu, and a list of recommended articles from other documents that are similar in content will be generated. You can then review the recommendations and choose to preview the articles or navigate to them.

This is just the beginning

We are excited about the future directions of using AI to make it easier to understand and use building codes. We believe that the current breakthroughs in Large Language Models, that powers ChatGPT and the new Bing Chat features, can not only improve the efficiency and quality of construction projects, but also promote the safety and sustainability of our built environment.

What do the building codes look like? How can we see the structure of the codes and compare them? Throughout the code harmonization efforts, to better align the national and provincial codes across Canada, can we see if the codes are becoming more or less aligned?

In the figure below, we compare the semantics between of all of the code sections for several Ontario and National codes. A key challenge in visualizing these similarities and differences, is determining how to flatten the hundreds of ways to interpret similarity into a usable two-dimensional image.

In the similarity map shown above, semantic overlaps, clusters, and gaps can be seen between the dots, or code sections. Each section has a complex “fingerprint” that is represented as a series of hundreds of numbers, also called a vector. Vectors are so central to AI concepts that the Vector Institute for Artificial Intelligence in the MaRS Discovery District is named after them! When the distance between these vectors is small, we take that to mean that there is high similarity. This is what drives our related document ranking.

We’re working on many ways to unleash powerful AI techniques on to the building code and regulatory world, so stay tuned… this is just the beginning.