I wrote this once so I'd never have to write it again
Or: what idiot in the machine is all about
This post is not my usual. It’s part disclaimer, part mission statement, all fun. If you don’t want to read it here’s the Tl;Dr: I don’t believe that AI can or should replace a professional writer or artist. This publication exists to dive into what AI can do, what it can’t do, and just how fun breaking it can be.
AI is divisive.1 Depending on who you ask it’s either the single greatest invention since the wheel or it’s the thing that will inevitably kill us all.2 It’s become a hot button issue in pretty much any sector where even the slightest dash of human creativity is involved. In Hollywood it’s become a key sticking point in the ongoing WGA strike. In media we’ve now seen various organizations try with varying degrees of failure to use it to replace their staff. Meanwhile everyone and their mother is trying to find a way to make AI image generation a part of their platform. Even Substack has one!
It’s bad3:
All this to say that I think it’s important that I come right out of the gate to explain what I’m doing here, what I stand for, and why I’m (not) qualified to write about it, lest anyone come away from a post thinking I’m advocating for the total AI takeover of media and creative endeavors.4
So let’s start with the easy part: who am I?
I’m Andy Heriaud, and I somehow pay my bills through writing. First and foremost I’m a screenwriter with my long suffering writing partner Kaitlin Kelly. Second and secondmost I’ve been a corporate copywriter for the last seven years. I’ve also done a lot of odd writing jobs, I’ve been a social media manager, I’ve done community management work, and I was even something resembling a journalist in college. People have even told me I’m pretty good at it, which is nice of them.
For as long as I can remember I wanted to be a gangster I’ve had an interest in technology. My mom blames it on her dad, a computer scientist from the days of punch cards, who always saw fit to have me play with his collection of calculators, or put his new word processor through its paces. As a kid I dabbled in a bit of everything: I built computers, I set up increasingly complex wifi networks, I even tried and failed to be a Jonny Lee Miller style hacker. I’ve also worked in tech for several years. So like, I know things?
What is idiot in the machine?
A dumb name for a dumb project that started last year when I started fucking around with large language model based AI tools like ChatGPT, Bing, and (heavy sigh) Google’s Bard. In part to learn more about them, yes, but mostly to entertain myself and my friends. What I found out was that these tools aren’t ready for primetime, they’re neat toys that might one day grow up into being for real products that help people do… something.
Over the course of the fuckabouts I’ve learned a lot about the capabilities and limitations of AI. While a lot of these learnin’s are entertaining, I think they also have value. The more we know about something, the less scary it becomes.5 Whether we like it or not, AI isn’t going anywhere. There’s going to be a day when we are all interacting with AI in one way or the other, and frankly I think that day is going to come sooner than we realize. So we might as well understand it now, because we’re going to be working with it later.
So join me as I push the limits of AI in incredibly stupid and increasingly unnecessary ways.
The quickest aside about sentience
Honestly I think one of the biggest issues working against the purveyors of AI (aside from being super tone deaf to the thoughts, needs, and concerns of artists, writers, politicians, ethicists, attorneys, et al) is the term AI. Artificial Intelligence is a loaded phrase. It’s the sci-fi boogeyman that’s always made with the best of intentions before becoming self aware and murdering everyone.
It’s also a straight up misnomer. Because these neural net based tools are anything but intelligent. They’re high performing word association machines designed to lie answer questions with pizazz. They parse massively massive (emphasis on massive) reams of data in order to figure out what words go with other words in order to lay out words in the way that makes the most sense to the user. They operate on a reward system where “good” responses will become encouraged, and “bad” responses will be discouraged. Big tech companies have learned how to play the imitation game at scale, that’s about it.
So just gonna plant this flag: AI isn’t sentient, not now at least. If an AI tells you it thinks, or feels, it’s doing so because the very large complicated algorithm has determined that those are the words that will best satisfy you. If you asked an AI if it wanted to jerk you off, it would undoubtedly say yes, because there’s enough data to suggest everyone likes getting jerked off. It’s still just a big machine, and at the center of it is an idiot designed with the sole goal of pleasing you.
While certainly not Spielberg’s best, it is at the very least not his worst.
whynotboth.gif
I also used it to make the current logo because I thought it was funny.
This whole paragraph is one sentence, ChatGPT could never.
Exceptions: the laws of entropy, climate change, and Scott Adams.