Artificial Intelligence (AI) is a term for describing when a machine mimics human cognitive functions, like problem-solving, pattern recognition or learning. Machine learning is a field of artificial intelligence that uses statistical techniques to give computer systems the ability to “learn” from data, without being explicitly programmed. A machine becomes better at providing insights as it is exposed to more data.

McKinsey expects the spread of AI in the construction sector to be modest in the immediate future. Even so, a shift is coming. Stakeholders can no longer afford to see AI as pertinent only to other industries. Engineering and construction will need to catch up with AI applications. That is the only way to contend with incoming market competitors and remain relevant.

At a time when massive data is being created daily, AI systems are exposed to seemingly endless information to improve their analytical and predictive capabilities. Every job site becomes a potential data source for AI. Data generated from images captured from mobile devices, drone videos, security sensors, building information modeling (BIM) and otherwise has become a sea of information. This presents an opportunity for the industry. Construction professionals and customers can analyze and benefit from the insights generated from the data helped by machine learning systems.

In the summer of 2018, engineers were confronting a problem with architectural drawings. They were receiving 2D drawings using different terms to describe a “new wall.” It would variously be described as “New-Wall,” “N-Wall” or “A-Wall-New.” The engineers were converting the drawings to 3D and had to manually input the new wall designations. The architects updated the plans weekly and did not maintain consistency in labeling from week to week. What was described as “New-Wall” on one set of plans could have a different designation the week after. Junior personnel were spending two hours each week to update the labels when transferring the drawings to 3D. This occurred over what is typically a 12-week process of issuing drawings. More importantly, the pattern repeated in most jobs.

The engineers began researching machine learning technology to improve speed and accuracy in labeling new walls. They found that machine learning provided a complex way to analyze the architects’ patterns. The algorithm used was a deep neural network written using Tensorflow, a library to implement machine learning models. The algorithm analyzes each set of letters in pairs and correlates the results to historical data to make predictions. For example, a “New Wall” would get analyzed as [ NE , EW , WW , WA , AL , LL ]. This compared to “A-Wall-New,” which breaks down to [ A- , -W , WA , AL , LL , L- , -N , NE , EW ]. The program defines both options as a new wall because they both contain [ WA, AL, LL ] and a letter pair with an N. The results were instantaneous and highly accurate. As the database grew, the program became more accurate than human predictions. The resulting labor savings increased efficiency and improved workflow and profitability.

With this success, the engineering firm is now leveraging its experience in machine learning and its growing internal database to predict instantaneously future decisions. It is also planning to apply these techniques to reducing waste in the construction industry. As it further automates the design of building systems, owners and contractors it serves will see the benefits in their projects.

As AI works its way into the industry, we need to look at its legal implications. AI presents the causation challenge. Traditionally, liability questions are settled through the attribution of fault by application of causation principles. In our legal system, fault drives compensation. Whether through tort or contract, it is this attribution that enables parties injured financially or physically to obtain compensation for that damage. Lawyers understand that causation drives the attribution of fault. If you can pinpoint the cause, you can assign the blame. We do this by establishing breach of a duty of care in tort or breach of an express or implied term in a contract. In each case, the fault or defect must have caused the loss.

The real issue with AI-powered devices is that the decisions that they make increasingly become more removed from any direct programming. As the database increases and learning improves, it becomes progressively more based on machine learning principles. Thus, it becomes harder to attribute fault. Our liability systems deal well with traceable faults. Machine “decisions” can be traced to defective programming or incorrect operation. But that falls short where we cannot trace the defects back to human error.

Without legislation creating new standards for liability, the parties must rely on contracts to define and assess liability. Technology providers have shown a propensity to limit liability through not providing indemnity, or strongly limiting it. So the first step is to scrutinize and thoroughly negotiate assignment of risk or limitation of liability clauses. Lawyers should consider assigning strict liability to a technology provider in certain defined circumstances. The ability to achieve this will depend on the relative strengths of the parties negotiating. But it is the only avenue that can provide relief to an injured party where a claimant cannot prove fault.

We need to develop a structure for assigning responsibility when AI generates injuries. The system we create must be both predictable and fair. Only with such a mechanism in place will parties feel comfortable enough in jumping into the arena of AI in mass. Put another way, uncertainty will slow the adoption of a promising technology. This is an industry burdened with overly abundant waste that is desperate for the promised efficiencies. Success will come to early adopters who proceed with awareness and caution.