Blog Viewer

Artificial Intelligence and Construction: A Look Ahead (AI Part 4)

By Gregory Ceton, CSI, CDT posted 02-11-2019 13:36


This article brings us to the conclusion of our series on artificial intelligence (AI). If you haven't read the other articles in this series, then the links are below. Review isn't necessary but may provide additional context or answer some nagging questions.

I had an opportunity at the beginning of the month to present on the “Future of Work” at the February Master Specifiers Retreat, and the ensuing discussion gave me a lot to think about. Presenting and learning being two-way streets, I discovered more about the daily work of folks in the AEC industry, and especially about the kind of tasks where intelligent assistance may be most valuable to practitioners.

AI: Expanding our reach

AI will provide flexible tools and resources that will enhance the capabilities of designers, project managers, contractors, and specifiers among others. There are many tedious and time-consuming tasks that do not require a trained and experienced professional but still need to be done. This type of task is the low-hanging fruit for sophisticated information management tools to address with less need for direct input and management by humans.

For example, the group at the Master Specifiers Retreat mentioned tracking, responding to, and resolving submittals as a task that takes a lot of time for construction team members. This is neither creative work nor does it demand the full faculties of a professional, especially the largely clerical tasks associated with tracking submittals. Nonetheless, because of contract requirements, and to ensure the timely and correct delivery of a project, it must be done. Since it is a form of detailed information recording and triaging that could be enabled by pattern recognition and other tasks that AI are good at, it is a strong contender for just the type of task to be addressed by an AI-powered tool.

Training and developing AI

A factor that could limit the development of AI tools will be the existence of datasets suitable for training the new AI tools to perform the tasks they are designed to do. Without suitable, sufficiently large, and well-structured datasets, it will be difficult to train an AI tool to perform at a professional level. Fortunately, we have an ever-increasing population of data in almost every area, but it remains possible that a chosen and particularly difficult problem lies in an area where little information has been digitized. This doesn’t make the creation of a tool in that area impossible, but it can increase cost and time to market while a dataset is created from analog resources and tested to ensure its suitability.

It is important always to remember that AI tools are developed by humans, they do not spring fully formed from a pond like an evolving frog. Their capabilities and scope are the direct result of human action to solve a particular problem. Also, as mentioned in an earlier article all AI currently in use is specialized, or narrow in scope. The current limits not only of AI, but also the processors which power them make the full substitution of humans performing work like design or project management unlikely any time soon. That said, the addition of AI tools will require a reshaping of work habits and workflows, as some tasks that may have taken up a significant amount of a professional’s time may now be nearly fully automated. Taking advantage of new tools will be key to surviving in a new environment where working without those tools will not be competitive.

Tomorrow’s tools today

With enough data to train them, AI can perform better, more cheaply, and faster than human professionals, but even so, they still lack the capacity of judgment. They can make strong recommendations for humans on the best course of action given the information presented, but the final choice of the appropriate path is and will remain in the hands of humans, at least for the near future.

AI will provide relief to a set of problems present now that will likely grow over time, including the challenge of accessing reliable information quickly and easily, increasingly short timelines to complete work, and budget and cost pressures and shortage of trained professionals to help share that work even if a firm could afford to hire them. AI-enabled tools and resources will help stretch every dollar and hour to its maximum and expand the capabilities of professionals in much the same way as the first levers and inclined planes expanded the capabilities of builders in the ancient world.