Advanced Computing in the Age of AI | Thursday, March 28, 2024

The Power of Babble 

When engineers talk about extrusion and tet meshing and FEA and so on, they're not trying to be pretentious — they're just speaking the language of engineering.

Back in school, one of my majors was Roman History. Upon learning this, people always ask some combination of the same three questions:Julius Caesar

  1. "So what did you plan to do with a degree in that?"
  2. "So why did they kill Julius Caesar, anyway?"
  3. "So why did the Roman Empire fall, anyway?"

Answer #1: Nothing. I liked it, and studied what interested me.

Answers #2 and #3: It's complicated.

I'm not dodging the question; I'm doing you a favor. You think you want to know, but you don't. You asked because people are naturally curious, but frankly, to explain, I'd have to explain my explanation. And then explain that. Let me give you a brief example.

"So why did they kill Julius Caesar, anyway?"

For several reasons, but chiefly the conspirators were worried about his effect on tradition, class and politics. Caesar's appointments rendered the Cursus Honorum irrelevant; if that continued indefinitely the very concept of dignitas would disappear, with the resultant loss of traditional auctoritas. Head count veterans wanted Ager Publicus reform, which he gave them, stirring memories of the Gracchi and adding fuel to the divisions among ordo equester, Patriciate, and Plebiate. While his dictatorship was theoretically viable in the eyes of mos maiorum, they feared he would become one of many, having already had a precedent in Sulla. That's the very short version.

How much of that made sense to you? Enough that you could follow it, maybe, but I bet a fair amount sounded like it was in Latin. BECAUSE IT WAS IN LATIN, and giving you an answer half in Latin means you're going to miss a bunch of stuff that I'd then have to explain, in English, and that would take way longer than you wanted. You just wanted to know why Caesar was killed.

It's complicated.

Torre di Babele by Flickr's aldoaldozI spend my days surrounded by engineers, doing my best to keep up with their talk of extrusion and tet meshing and FEA and so on, but honestly, a ton of it is beyond me. They're not trying to be pretentious — they're just speaking the language of Engineering, as I can give a short answer to a complex question provided I'm allowed to speak in the language of Roman History. Stuff like "Cursus Honorum" and "ordo equester," I know what those things are. Not just the translations, I know what they mean, in the context of history and culture. Without explaining all of that, I can't give a clear answer to why Caesar was killed.

To give a non-expert a comprehensive and comprehensible answer would take a long time in any area of expertise. Think about it: doctors, lawyers, plumbers — they all have dialects that make their communication and understanding efficient.

If you don't understand the dialect, you're out of luck. In part because asking is, frankly, embarrassing, and usually pointless. I say, "okay, wait, what's a tet mesh?" and the engineer says "it's a model from design that defines a part," as if it's obvious and only a stupid person would ask.

But then I'm like… Why are the model and the design different things? Isn't a model already a mesh? Is a mesh not a wire frame? And why do you need a mesh? And what's a part? Is a part a washer? Is a part a carburetor? I'm not ignorant — but I am ignorant of the language you're speaking. And the more I ask for translation, the stupider I feel, and the less likely I am to be on your side.

We have to be sensitive to this when introducing a concept as revelatory as digital modeling and simulation. The small and medium manufacturers who need this tool the most struggle with the how, and the why, and the what it does. I've observed this even in people who want to be on board, and I sympathize, because often the answers they receive make no more sense than whatever drew a question in the first place.

"You can upload a job and mesh it, do an FEA, and material analysis."

First of all: what?

Second, these small- and medium-sized manufacturers (who we call SMMs, though that's the Language of Digital Manufacturing so be careful of tossing it around and assuming everyone understands), do they do that now, in house? If not, do they need it? If so, what will it do for them? And don't say it'll allow them to upload a job and mesh it.

We talk about digital manufacturing, and SMMs, and meshing, and jobs, and all this stuff like people are expected to know what it means. Of all the barriers modeling and simulation (M&S — wait, you didn't know that?) face, one of the biggest is that its proponents, though well-meaning, are the equivalent of pioneers landing on the shores of North America and sharing no common language with the natives.

We must make an effort to understand, and to be understood. And so far, from what I've observed, we're not. We're making it worse.

Remember, M&S — seen as part of the whole that is digital manufacturing — is a revolutionary leap forward. It's like going from the telegraph straight to email. How do you convince users who are still thinking in terms of Morse code, and going to the telegraph office, and STOP at the end of sentences, and paying by the word?

Not like this: "A local or Web-based client sends out on SMTP Port 25, communicates with the DNS server, then reassembles packets through POP or IMAP protocols on the other client."

And yet that happens all the time when we try to communicate the value of M&S.

Do you know why the Procter & Gamble Pringles M&S story is so popular? Because it's in English. The chips were flying off conveyors. How to stop that without changing the signature shape of the chip? Use a new tool that makes tiny aerodynamic tweaks, small enough that no one will see a difference. This wouldn't be possible without the new tool, because the old way just can't make such minute adjustments or effectively test them.

We must speak in allegory, in metaphor, and in results. Tell the window glass maker how much thinner, lighter, stronger, and cheaper he can make his product if he leverages M&S. Tell the cargo hook manufacturer how M&S can help predict unexpected stresses well beyond what he's already prototyped. Tell the air conditioner manufacturer how this subtle change in fan shape will save homeowners 65 percent on their summer power bills. To prove it, show them case studies.

And when they ask how, have them give you a problem or challenge or new product or scenario or something, and then sit them down and walk them through the process from beginning to end, and do it in a language they understand. Explain how their own people already have enough knowledge to take advantage of M&S, we're just providing access. And when they ask how much, remind them that they'll only pay for what they use: no up-front investment, no weeks of sending employees offsite for training, no hardware and software purchases.

Time and again I've seen SMMs frightened when they should be exalted, confused when they should be jotting down ways they can already see themselves using this, reluctant when they should be asking where to sign up. It's not because SMMs refuse to get it, it's because it's complicated.

EnterpriseAI