Generative artificial intelligence (A.I.), predictive A.I., machine learning and neural net are terms often associated with sci-fi movies such as “Star Trek” and “The Terminator,” but they have many common uses in businesses today. If I were to ask when A.I. made its debut, what would you say? Take a minute and think back. The answer may surprise you.

Generative A.I. is used to create unique content, including pictures or phrases. Predictive A.I. is used to predict future events based on historical data. Machine learning combines the two and contains a “learning” phase. Each has had its own progression, but by definition, A.I. was born in 1964 at the Massachusetts Institute of Technology with the development of a chatbot named ELIZA.

Once a computer had the ability to store data and “learn” based on the result of an input, A.I. was a reality. But that was the technology’s infancy. The first mainstream mention of A.I. wasn’t until 1997 when an IBM supercomputer known as Deep Blue beat chess grandmaster Garry Kasparov in 19 moves. Chess is a difficult strategy game. Every move opens the opponent up to a counter­move, which leads to a move from the challenger, and so on. The possibilities seem endless from the opening to the final declaration of “checkmate.”

With real-time decision-making and adaptations based on live actions, modern machine learning was born. But what exactly is machine learning? It is a computer’s ability to make decisions and adapt on its own with limited instruction during its training phase. Eventually, it does not need explicit instructions to perform a task, such as playing a game of chess. This imitation of human behavior, or “learning,” is what makes up A.I.

A.I. has grown from playing chess to being incorporated in more areas today than you may realize. A streaming service may suggest “Escape From L.A.” after you watch “Escape From New York.” Your audiobook app may recommend a few related books after you finish the one you are listening to. You are using A.I. when you ask Siri for directions to your favorite restaurant and it auto-routes you because of traffic. When you talk to your friends about going fishing and see ads for fishing poles the next time you hop on the internet on your phone, that’s A.I., too (although a somewhat intrusive example).

Artificial intelligence timeline produced by ChatGPT. Mistakes marked in pink by Electrical Contractor. Always use your human eye to verify and edit anything A.I. generates in your name.

Generative A.I.

Let’s focus on how each type of A.I. affects the electrical contractor. Toby Mitchell from Classic Electric and Consulting, Nipomo, Calif., began incorporating ChatGPT into his business about a year ago and is finding new ways to implement it on a daily basis.

“We have replaced our safety software by using A.I.,” Mitchell said. “With the software suggesting talking points, it would give us safety topics that did not necessarily pertain to our area, for example, hypothermia in July living in California. By using A.I. to create our talking points, we can tailor them to topics that directly relate to challenges that we are having. A byproduct of this was more participation during safety meetings because the topics apply to actual day-to-day challenges of our field personnel; for example, safety topics on driving in the fog.”

Safety topics and toolbox talks are just one way he is using this technology. 

“We use A.I. to proofread emails before they are sent out[.] If we are having a hard time starting an email, we can throw out some talking points and use the key words, ‘draft a two-paragraph email about …’ It will respond with a two-paragraph rough draft. It will not give you a finished product, but it at least gets you a starting point that you can build from. 

“This alone has improved our productivity; we no longer have to ask someone to stop what they are doing to come proofread a document,” he said.

Using questions and phrases to create unique responses is an example of generative A.I.

Predictive A.I.

Tracking the condition and performance of electrical equipment such as switchgear and transformers uses predictive A.I. to spot abnormalities that may indicate potential faults, which could cause fires or shock hazards if not addressed in a timely manner. By preemptively notifying facility supervisors automatically, potential hazards can be addressed before failure.

Predictive A.I. can be used in the smart microgrid sector where energy-forecasting A.I. uses automated processes to predict how much power a building will need over the next day by looking at past energy use and weather conditions. This helps to diversify the power requirements more wisely and ultimately cut costs. 

In conjunction with machine learning, this type of A.I. can be used to evaluate asset management, which in turn cuts down on false alarms and unwanted data, making it easier for maintenance teams to respond quickly to actual situations. Together, these tools help keep smart buildings and microgrids running smoothly and at peak efficiency.

Aaron Szymanski, co-founder and chief product officer of Toronto-based Augmenta, explained how the company is using A.I. to affect the electrical construction industry, including with the introduction of auto-­routing conduit during the coordination phase using Autodesk Revit building information modeling (BIM) software files tailored to specific contractor workflows. By applying rules derived from thousands of hours of contractor interviews, they were able to gather information on best practices, means and methods, as well as code compliance and constructability.

The result is an A.I.-based system that uses Revit to automatically route and coordinate conduit in a fraction of the time it would take a human modeler to perform the same task. It took years to get Augmenta to this point, but it all started with the development of the auto-router, followed by an analysis of the routing, conduit configuration, then constructability review. The A.I. takes all these factors into account to provide an automatically routed conduit path, considering factors such as go/no-go zones, conduit lengths and different routing paths based on bend, distance and price. This process is achieved through hybrid A.I. machine learning and a rules-based algorithmic system.

Augmenta, which is geared toward larger contractors, uses generative A.I. to auto-route in the BIM software based on input from users, usually in the form of specifications and submittals. It then uses generative A.I. to try to auto-route future conduit based on the results and inputs from previous jobs.

AECInspire, Newark, Calif., is an A.I.-based software electricians can use to detail electrical systems to get the bill of material information for their projects. The AECInspire team discussed an example of A.I. in its machine learning capabilities and how the company is finalizing the A.I.’s “training” through predictive A.I. with user input. 

The current A.I. can scan through the document and find symbols, recognize what they are and log them with minimal user input in certain situations where the symbol may not be clear. This system already reduces the dreaded “spots and dots” process from hours or days into minutes. In the chart above are some sample results.

AECInspire is geared toward smaller electrical contractors and gives companies the ability to digitally create a bill of materials, which is usually something only big contractors using BIM can do. It is equivalent to an A.I.-generated, entry-level BIM. A contractor with five electricians can access some of the same digital tools as a contractor with 100 electricians and a full BIM team. It also drastically decreases the time to estimate a job. As seen in the table, while it would take at least an hour to estimate 17 sheets manually using a PDF, AECInspire’s A.I. can do it in under 6 minutes with 98% accuracy.

Electricians can use AECInspire software to scan a document and find symbols, recognize what they are and log them with minimal user input. This chart shows sample results.

Machine learning

With an anticipated update release in 2024, AECInspire is looking at advancing to the point that the user input to search for symbols will be gone, and the A.I. will be able to recognize electrical symbols as the files are uploaded. For example, as contract documents are being uploaded, the A.I. is scanning and determining what symbols are receptacles or power data devices, as well as determining lighting types based on their symbol. Once it completes its initial scan, the A.I. would provide a list to confirm any symbols in question, then “learn” based on the user’s clarification of the symbols.

How ECs can use A.I.

How is this affecting electrical contractors? Using the example of predictive A.I. and machine learning, you would be able to upload and count 16 unique symbols on 17 pages, with 2,372 devices identified in under 6 minutes with a 98% accuracy. Using this data, as opposed to the spots and dots method, is just one way electrical contractors can save time with A.I. when used in an estimating workflow.

A.I. has been a buzzword for many years, but it has rapidly moved to implementation. If you were to compare the progression of A.I. to that of human development, the predictive and generative phases while being trained through machine learning would be equivalent to our academic years when we are being taught in school.

The implementation of a neural net—the ability to self-learn—takes machine learning to the next level, called deep learning. Deep learning would be the equivalent of graduating from school and learning through life experience and trial and error. 

These are examples of A.I. now, and the technology is advancing exponentially. With robotics and A.I., the construction industry is preparing for a technology boom. If you doubt that A.I. plays a considerable role in that boom, consider what a robot such as Boston Dynamics’ “Atlas” or Trimble’s “Spot” would be without it. Still, electrical work will always require the human touch.

stock.adobe.com / Summit Art Creations / Ilmi Jared Christman / Bonotom Studios Inc.