AI is Like a Concrete Chainsaw: Powered From a Distance, Environmentally Deleterious, and Only Good for Specific Tasks

Dave DeChant slices 100+ year-old concrete with a hydraulically-powered specialty chainsaw. In many ways, much like an AI application. You can see on the right where we cut the wall away, then knocked it over to cut into movable pieces.

My first career (way before my work in law and technology) was in real estate development and property management. Starting in the 1990s, my family’s business was buying, renovating, renting, and selling buildings. We were bootstrapped in the beginning and earned (actual) sweat equity along the way.

My dad (shown above wielding our chainsaw) was a foundryman in his previous career, melting and pouring molten metal into molds. He often quipped “A bad day in real estate is better than a good day in a foundry”, but as a foundryman he carried, and instilled in me, a “get it done no matter what” mentality. That’s how we wound up with our concrete chainsaw, an expensive (when new) specialty construction tool.

We were remodeling an old house in Chicago’s Bucktown neighborhood, and needed to remove a concrete foundation wall more than a foot thick and over 100 years old. As I learned, the chemical process that hardens concrete never actually stops- so old concrete is really, really hard. We tried everything we knew to knock down that wall, using drills, hammers, powered chisels, everything. Even the gas-powered concrete circular saw we had wasn’t going to cut it.

A classic Chicago worker’s cottage, originally built 1870’s (pictured post remodel). We just left the red gingerbread roof trim alone.

Knowing we had other projects with even thicker walls to work on, we did what any rational person would do: buy a used ICS 890F4 hydraulically-powered concrete chainsaw with a 24″ bar on eBay. With hydraulic power, you can have a big, powerful, gasoline-fueled stationary “power pack” running somewhere else on the job site, and connect it to the chainsaw via hoses. The power pack pressurizes oil to 2,500 PSI (your car tires are inflated to 35 PSI) and pushes it through special hoses to the saw. This allows us to leverage an enormous amount of power in a hand-carried tool. At about 14 horsepower, a saw would weigh 100+ pounds if running on electricity or a gas engine mounted on the tool itself. Instead, we could carry it around, set up the hoses and start cutting. Added plus: no nasty gas fumes to make us sick while running the saw.

The powerpack burns gas and stinks up the yard, while we get all the power made by the pack pushed into a relatively small hand-portable tool. On a later job, we ran the powerpack in our work van since we had no other space to run it (van interior smelled like gas for months afterwards).

It’s the remote power aspect of our saw that stuck me as very similar to AI tools. With my phone and a ChatGPT account, I can leverage incredible amounts of power based in data centers and server farms spread across the country, which not only consume resources to deliver what I ask for, but have already consumed vast quantities of resources just to get built and trained. Beyond that, the remote nature of the power source means less local control over it; whether it’s a data center or powerpack, it’s up and running wastefully waiting for you to engage with it.

AI tools like ChatGPT not only consume resources (somewhere you aren’t) to run, but also to build and train. The power to feed those processes is generated (somewhere else you aren’t), emitting greenhouse gasses, toxins, and waste.

I’ve heard that every ChatGPT prompt request consumes the resources needed to fill and transport a disposable bottle of water. Here’s a few more numbers:

The other common trait the two tools share is their narrow applicability to specific tasks. A concrete chainsaw does one thing really well (you can’t cut a tree down with it), and you need to be within your hoses’ length of the powerpack to run. You need an internet connection to run ChatGPT, and it’s good at language synthesis, but not much else. This is why I see so many professionals out there mocking ChatGPT results when tasked to perform a feat it wasn’t meant to do (like give legal opinions and case citations).

The end result of our remodeling efforts. In the opening video, my dad is standing where the potted plant is, and the wall we cut away stood where the ottoman is now. I fabricated the custom stairs and woodwork from large beams found on site.

I could tell you I’ve got the most powerful and user-friendly chainsaw ever made, but if you try to cut a log down into firewood with it, you’ll be disappointed.

We need to clarify how we see the complex parts holding modern business together.
We spoke to Eric DeChant, a Legal Engineer at xMentium and the Founder of the American Society of Legal Engineers. We speak
Scroll to Top

Discover more from The American Society of Legal Engineers

Subscribe now to keep reading and get access to the full archive.

Continue reading

Enter Email for Access

Join Our Newsletter