Featured Image

CAM GPT: Part 1

Between using AI for homework, wedding vows, and asking for raises, society can hardly be faulted for overlooking one of the most important areas the tech can be applied to: cutting things out of metal. AMT tech analyst Stephen LaMarca wants to amend that.
Jun 05, 2023

Unless you’ve been living under a rock lately, there’s been a ton of hubbub around artificial intelligence (AI), the AI chatbot ChatGPT, and the AI language model GPT4, as well as discussions regarding accessible AIs (read: web-based and open to public use) after recent “hello world” debuts. People use this tech for everything – doing their homework, programming PLCs, writing their wedding vows … My fiancee even used ChatGPT to negotiate a raise via email – and it worked! Some of the buzz is positive, and some is negative – like really negative. When it comes to AI, there will always be detractors who think “Terminator” machines will come for us. As far as I know (at least since starting this project in February), nobody’s attempted to use ChatGPT to program a machine tool, much less arm it with plasma weapons. So, since I hadn’t used ChatGPT yet, I figured: Why not try? (The machine tool part, not the arming it part.) 

What’s Going On?

You see, Millennials like me Google everything. If my car throws a check engine light, I’ll Google the OBDII fault code. If Google says the mass air flow sensor is probably just dirty, I’m going to YouTube to see how to clean it. Sure, everybody does this now, but we were the first, and we’re exceptional at it. Why? Because our teachers told us, “You won’t always have a calculator!” How’d that work out? Not only do I always have a calculator, I can also access almost every piece of music ever created (and its misheard lyrics) in fractions of a second in high fidelity! My point is that no matter where you are, you’ll always have a calculator – and now it can access AI!

Come Out of the Fever Dream, Please.

Sorry. All I’m saying is I think AI is going to be bigger than the internet. It’s more on the magnitude of discovering fire. Let me show you what I asked ChatGPT.

The image below was my first interaction with ChatGPT. I nearly teared up. 

When I started at AMT, one of the first things I attempted was to program our 5-axis CNC (with zero manufacturing background, mind you). Unfortunately, the machine tool manufacturer (Pocket NC, now Penta Machine) used clunky software. I don’t want to put the actual company on blast, but they promised their product was the easiest all-in-one CAD/CAM solution ever. Long story short: After seven years in the industry, the polygraph test determined that was a lie. Over the years, I’ve spoken to a ton of software developers and startups dedicated to making the CAM process less aneurysm-inducing. Not much luck. 

Enter ChatGPT.

Lo and behold, after just two dialogs with OpenAI’s Chat (that’s what I call it), this image below is what I got back.

What Kind of Feedback Have You Received?

Well, I tell you what, I got a few doubters and a few supporters. But I will say this: Contrary to what some skeptics might think, I’m not going to blindly upload this to a machine and run it. No machinist worth their salt would do that. Ever. No matter the source of the program. Seriously, have you ever heard a production machine tool operator speak highly of their CAM/manufacturing engineer? They trust them about as far as they can throw them. You sim first; then, if it doesn’t go back to engineering with immediate issues to resolve, you upload it for an “air cut” (no tools, workholding, or stock material). If by some miracle the program makes it that far on the first draft, then the machinist may cut wax to cover their bum and be sure. The planets would have to align for that to happen, though, and I don’t foresee a CAM engineer tearing space-time any time soon.

What Can We Expect Next?

As I stated above, I’d be dumb to just throw this program into the machine and press play. That’s begging for an “epic machine crash.” Anyway, next step: simulation. A keen eye might see that this program gets bounced faster than a new version of anything by Microsoft needing to be patched. See you next time! 

If you have any questions about this information, please contact Stephen at slamarca@AMTonline.org. For more LaMarkable content, stream “Road Trippin’ with Steve” now on IMTS.com/rtws.

Stephen LaMarca
Technology Analyst
Recent technology News
Finally! Industrial-grade Arduino. DOE raises $45 million with AM hand. How do AM shoes wear and tear? Low-emissions LEGOs, "Alexa, buy Anthropic."
Mass-produced heroes. Apple to 3D print each series 9 case. Wall Street bets on robotics. Build a deep learning rig on a budget. Manufacturing refuge.
CDFAM23 showed that household brands such as New Balance, HP, Carbon, and Adidas are using computational design tools today and demanding tomorrow’s advanced software tools to maintain a competitive advantage for the future of footwear.
The inaugural Computational Design Symposium established the state of the art in computational design and generative artificial intelligence software for engineering and manufacturing applications, featuring speakers from NASA, MIT, and more.
The CDFAM 2023 symposium in New York City explored computational design’s latest trends. Twikit, Mode Lab, and Synera showcased scalable design solutions that are faster, cost-effective, and improve customer satisfaction through mass customization....
Similar News
By Stephen LaMarca | Sep 29, 2023

Finally! Industrial-grade Arduino. DOE raises $45 million with AM hand. How do AM shoes wear and tear? Low-emissions LEGOs, "Alexa, buy Anthropic."

6 min
By Stephen LaMarca | Sep 22, 2023

Mass-produced heroes. Apple to 3D print each series 9 case. Wall Street bets on robotics. Build a deep learning rig on a budget. Manufacturing refuge.

6 min
By Benjamin Moses | Sep 20, 2023

Episode 102: Ben and Steve share stories and takeaways from recent travel, including Oak Ridge National Lab, FABTECH, and the Outer Banks. Benjamin shares NVIDIA’s take on the impact of generative AI in robotics ...

41 min