All I’m Sure I Know About AI

Welcome to our February issue, or (as I like to call it in editorial meetings) our Big Tech issue. While the physics of plumbing will never change (water wants to flow downhill, steam wants to condense), the technology surrounding the heating, delivery and treatment of water is constantly evolving.

We have seen some amazing innovations sweep the industry. BIM and Virtual Design and Construction (VDC) have changed all phases of construction. Robotic automation and prefabrication techniques have realized incredible efficiencies. Connected “smart” devices have revolutionized how mechanical systems are controlled, commissioned and maintained.  

But the big story these days is Artificial Intelligence—and not just for the mechanical contracting industry, but for practically every industry. In my work as a writer and editor I have never had to set up a smart circulator (well, maybe once at a product demo), but I’m now working with AI almost every day. (Our feature this month is on AI and its role in contracting—it starts on pg. 18.)

I’m old enough to remember the early days of the Internet—when going online involved a dial-up modem—and now feels a lot like then. There’s the same excitement over the possibilities, the same trepidation over the pitfalls. It seems like for every story about AI making a miracle diagnosis to save a patient’s life there’s another one where AI-generated advice causes a person to nearly poison themself.

Between our hopes and our fears and our greed (because let’s face it, there’s a lot of people making a lot of money with and off of AI) there’s a lot of conflicting information. Here’s all I know for sure.

Artificial Intelligence isn’t really “intelligent.” True AI (what the researchers refer to as Artificial General Intelligence) is still the stuff of speculation and science fiction. The Large Language Models most of us are calling AIs aren’t capable of reasoning, complex problem solving, or learning in the way that human minds are. ChatGPT isn’t thinking any more than a calculator is when you input a problem—it’s just a machine obeying a set of instructions (albeit a very long, complicated set of instructions).

There are a few things this implies. First, that we won’t be facing a Wargames or Terminator style robot takeover anytime soon. These are, however, very powerful, easily accessible tools, and not everybody using them has the purest intentions. Instead of a nuclear apocalypse, some have speculated how bad actors could use AI to create an “infopocalypse,” where a constant stream of AI-generated articles, images and videos leave the public without any reliable sources of news.

Second, just as with any computer program you might be using, remember GIGO: Garbage In = Garbage Out. Since AI can’t think for itself, it relies on the data it’s given (or can find) and the instructions it’s given for using that data—the prompt. (Melvin Newman, writing in this month’s feature, gives some advice for creating better AI prompts.) Bad data and/or bad prompts will lead to bad results.

Third, AIs can mess up. Their fallibility might be the most human thing about them. These machines are so eager to please their users that they will invent facts, quotes, entire reference libraries out of thin air. Always check your outputs before sending them to co-workers, clients or customers. AI is for the first draft, not the finished product.

And lastly, AI is here to stay. For good and bad, there is no putting the genie back in the bottle. So get some experience using these powerful tools. If you’re unsure where to start you can describe your business and its current pain points to an AI and ask for suggestions, advice and training. There is no question you can ask it won’t try its best to answer—just review those answers with a little human common sense.