Editorial: A word about ChatGPT

Our cover story this month is about autonomous compaction technology, which mostly concerns asphalt and trench rollers because they seem to be the closest to reality. When and if all the bugs are ironed out and the cost is reasonable, we should be able to rent our customers a roller that does the job itself without an operator dedicated to it. That’s a pretty sweet savings for a contractor, and a homeowner or farmer would get a better job done than they can do themselves. 

That’s the promise of AI as applied to construction equipment. There are other applications related to collecting data to improve logistics, scheduling, maintenance, safety, work execution and more. I’m sure they are very exciting but they are probably not being deployed by anyone other than the largest players. But a machine that does the work itself while you walk away is an advantage anyone can appreciate. 

Even automation’s biggest proponents admit full automation of construction processes, even simple ones like rolling, is probably still 10 years off. But AI has another potential application that is much closer to implementation: creating magazines like this one. 

At Christmas last year my family sat around the dinner table after the meal while my brother played with ChatGPT on his phone. He had the AI compose songs about each of us, complete with chords, then picked them out on his guitar. The results were hilarious, but also astonishing and, for a content creator like me, a bit unnerving. All he needed to do was put in a simple prompt – “Compose a song with lyrics and chords about Charlotte’s eyes in the style of John Lennon,” for instance – and seconds later an apparently completely original composition would come out. The songs certainly weren’t great but sometimes they were no worse than a lot of the crap you hear out there. 

I haven’t played with it extensively but hypothetically it should be possible to generate a
feature-length article on some general information topic this way. “ChatGPT, write me an article in the style of Canadian Rental Service about autonomous compaction technology,” for instance. If it could generate something half decent (so far, I haven’t found that it can) it would save me hours of time and/or our magazine hundreds of dollars in freelance writing costs. There it is: the fear that a new technology will cost jobs, brought for the first time to the creative sector. It’s here and it’s serious; the recently ended Hollywood writers strike was largely about getting protection from this threat.

 Our jobs here at the magazine seem safe enough for the present. ChatGPT still can’t interview Canadian rental store owners, understand what they have to say and write an article about them that captures what’s interesting about them and their businesses. When I put in the prompt above to write an article like our cover story, I get something generic and shallow with no new information or insight. I hope you agree I can still do better. 

Finally, there’s a deeper point. Whether we are conscious of it or not, every time we read something we carry with us the assumption that a human being stands behind the words. How we receive those words and the way we react to them depends on our assumption that the person giving them to us knew what they meant and intended to convey the message on the page. AIs don’t have even this basic level of awareness. They have no idea what they are saying or if it’s right or wrong in any larger context. To present their work to you in the same way we present a human writer’s work would be fundamentally dishonest. 

That’s why I commit to you here that you won’t read AI-generated copy in Canadian Rental Service without attribution as such. And I’ll make every effort to make sure AI-generated images are flagged, too (though that’s a harder one with so many images being contributed by third parties). It’s important for you to be confident that what you read here is the product of the real insight and judgement of real human beings.