Who's Training Who?

If you're like me, you've recently had to have some "training" at work concerning AI.

I'm not talking about the training my son had when he first arrived at grad school, where he learned how to spot AI generated work that students would pass off as their own, but the basics of using AI to "improve workflows" and to "increase productivity".

Oh yes. I believe it's a sign of the apocalypse that when your employer wants you to learn something, said something is about to implode.*

Anyhoo, I was thinking about all of that training over the weekend when this YouTube video dropped in my lap:

Now, I'll be up front is that I find Brandon Sanderson likeable and engaging as a person, or at least how he presents himself online, but I'm not the biggest fan of his work. I liked Mistborn and The Well of Ascension well enough, but the concepts that he'd put forth in The Stormlight Archive really don't appeal to me. We also disagree on the genius behind Robert Jordan's Wheel of Time series, because he really thought those middle books --where I and a bunch of other readers finally had enough-- were much better than we gave it credit for.**

I'm not personally acquainted with Brandon either, but I have enjoyed his class on writing SF&F. There's actually two versions of the class on YouTube, last years and one about 3-4 years ago. To be honest, both are worth watching if you want to get a feel for the nuts and bolts (and the business) of writing in my favorite genre.

All that being said, I really found his keynote address from the Tailored Realities release event at Dragonsteel Nexus 2025*** quite interesting, and I found I did agree with it on a lot of points. I'm more of a Original Series Star Trek fan myself, but I do appreciate his comparisons of Data and his ongoing desire to become human in the series as a reference point to what separates current levels of pseudo AI from what most people would consider art. But more than anything else, what resonated with me was that art is a transformational process: the creation of art also changes us, and the more we remove ourselves from that process by the insertion of AI into the mix the less we are transformed as well. 

That doesn't mean that we can't be transformed by creating AI art, but our isolation from the creative process makes it harder for us to be transformed. After my "training", I've compared what's known as vibe coding as basically what a non-technical boss thinks that coders do when they provide a design document. Instead of taking a framework created by your prompt and then coding in all of the details, vibe coder instructions are to instead "refine your prompts so that you get what you want that way". To me, that's equivalent of the old Pointy-Haired Boss from Dilbert trying to tell the engineers how to do their job.

Holy crap. This was in 1995? Yikes.

It also gives new engineers and coders a false impression of how to write code. You need to learn to fail before you can succeed, and vibe coding bypasses all of that knowledge by eliminating the skill development process. It's great if you're an experienced coder, you're in a rush, and you already know what you want, but to do regular work on an ongoing basis? You're reducing the code writing process to "just give it a few words and let it do the job". 

I suppose I could boil down a lot of the "promise" of AI to this: it lets MBAs think they can be engineers or scientists, without them realizing that the their own job could be replaced as well.

*Back in 2000, my then employer's CEO called all of the development staff together for an all-hands meeting. The late 90s, if you'll recall, were the high-flying days of the OG Dot Com boom, and it was all "internet this" and "internet that". So, when we were brought together despite the push to get the current version of our software out the door, we joked half-heartedly that the CEO was going to announce we were going to be an internet company.

Lo and behold, he actually did just that. We were going to put our mid-range CAD/CAM/CAE software, which was so hefty that it could barely run on the best Windows XT servers at the time, on the internet as a browser-based product.

We were all stunned. The network throughput on the internet wasn't up to the task, and more importantly neither were the browsers themselves. It was an idea 20 years ahead of its time, perhaps, but it was also a harbinger that within the year the Dot Com bubble burst and the stock market imploded.

**I also thought that Robert Jordan could have used an editor who reined some of his worst impulses in, such as his tendency to overdo it on the language and description. I mean, I'm not the greatest writer of description in the world, but I really do think that RJ was really just padding his page count at times.

***In case you wondered where the Keynote part of the title is about; I know I sure did. Here's a blog post of the speech itself from Brandon's website.