Generative AI is useless for the construction industry; here's what works

Champ Suthipongchai, General Partner at Creative Ventures

Startup drama has occupied much of tech news over the last two months. It’s been more tabloid than informative, from SBF’s trails to Sam Altman’s departure. Perhaps this gossip is just the tech world’s British monarchy.

Still, the interest is valid in some regards. After all, ChatGPT sparked a new genre of AI and surprised the world. So, let’s take a step back from the name-dropping and scandal for a second and really look at what generative AI is and how it is (or isn’t) beneficial for the construction industry.

Generative AI is useful if it doesn’t need to be right. GenAI is a “Stupid Idea Generator,” which sums it up.

Before generative AI came about, regular AI (aka discriminative AI) was about distinguishing different types of information inputs. That is, the AI can tell A versus B to assist with predicting anything, from whether a patient might have cancer to helping robots orient their arm to pick up an asymmetric object.

Regular AI can be considered humanity’s “left brain,” essentially. This makes Generative AI humanity’s “right brain.” It is a creative form of AI that comes up with high-quality content with very little directive. The content comes in various formats: text, images, videos, songs, poems, and, of course, the answer to homework.

The problem is that, much like new and unexamined ideas, they are often wrong. The generated content appears very useful when subjective and doesn’t have to be “right.”

So, if you’re looking for a new logo or are just in the business of subjectivity, Gen AI is the best thing since sliced bread. Music composition, content creation, design, and talking to your customers have never been more effortless.

But if you’re in the business of having to be right, then Gen AI is just some new fad. This is the reality for the construction industry, where the difference between right and wrong costs millions—if not billions —not to mention the potential loss of life.

Construction AI needs to be bound by the laws of physics.

I’ll be damn if even a single commercial real estate design created by generative AI is implementable. (But it won’t be because generative AI does not follow the law of physics.)

"GenAI is a “Stupid Idea Generator,” which sums it up."

On the other hand, discriminative AI is much more useful in unlocking construction bottlenecks by encoding AI’s deep learning statistical analysis with a physical world model. Construction AI needs to be practical.

What if construction scheduling took 2 minutes instead of 2 weeks?

If we’re lucky, a series of Post-it notes scribbled on over two weeks gets translated into a Microsoft Project and a construction schedule is born. And until there are serious delays and changes, no one will bother making the slightest change to that schedule.

With the advent of discriminative AI, ALICE Technology created the world’s first AI construction scheduling software that allows any contractor to churn out a schedule in mere minutes. The contractor can optimize cost and time and make decisions based on different resource constraints, such as labor, materials, and equipment availabilities.

ALICE is useful for one simple reason: the schedule that is generated actually works—and does so much faster than a master scheduler.

Decision automation is where AI is useful for the construction industry.

From construction scheduling to budget forecasting, AI has the potential to enhance productivity by 10x and drastically reduce time and cost overruns that plague the industry. But all these use cases are about automating decision-making in complex environments, much like construction sites.

Many of these problems center around resource planning but do so in an integrative way. Fusing disparate information together is how an AI will shine.

The holy grail of construction AI is one that takes CAD or BIM design and spits out accurate cost projection and schedule since the pre-con phase, and in doing so, optimze between the two from the design stage to avoid design change at later stages.

In the construction phase, the same AI captures construction progress and makes adjustments to the schedule based on different constraints. In the post-con phase, the AI becomes a single source of truth and directs maintenance efforts to the right place in real-time.

When AI can fuse design with cost management and scheduling with progress tracking—and then tell us what the most optimal path and outcome are—we will truly have the ultimate AI for the construction industry.