Books and Ghosts

Patience and Chivalry commission for @aporeticelenchus, with a little Fitz as a bonus because it made me feel nostalgic :D Thank you!

However, I do think this case is great example of what happens when smart people decide they can use these systems “responsibly.”

Let’s put aside the plagiarism and the energy use for a second (not that these don’t matter, but they’re not relevant to my point at the moment) and pretend these systems were made with entirely properly obtained info and ran on only solar, but otherwise were entirely the same. If you’re a person, you might think “okay, I’m going to only use this to come up with ideas for my next paper: I’m going to do all the research myself and the writing, I’m just gonna brainstorm topics.”

The thing is, even if you’re trying to limit use to just “brainstorming” the chatbot isn’t going to limit itself to just brainstorming. You might ask it “what should I write about involving Greek history” and it may suggest “the plague of Athens” but it’s also going to suggest some “basic info” on the plague of Athens that you might end up picking up and deciding your topic is. It’s going to provide you with “research.” And even if you decide to check that research yourself, that’s still setting you down a specific path you might not have picked organically.

And speaking of double checking its “research”, fact checking is a time heavy process. Not only now do you have to read the original text the bot provided, but you also have to compare it line by line with the summary the bot provided. That’s more time than it would have taken to just read the article by itself. And with any time heavy task, people are going to get lazy. And in original research being lazy might mean skimming a paper or an abstract, chatbots allow you to be lazy without looking at the article at all.

And soon enough, despite telling yourself you were only going to use it for brainstorming, you have offloaded topic generation, research and eventually even drafting because that’s what these bots are intended to do. They are not meant to serve as research tools or help brainstorm or serve as a bouncing off board: they are built to encourage maximum engagement. And that means if you go at them with the best of ethical intentions, they are built to tempt you to use them as much as possible, regardless of whatever line you drew in the sand for yourself.

Obvs, Hank’s situation is not the same as the scenario I laid out but I think the snippet above gets back to my main point which is these tools are built to encourage interaction and that’s a hard thing to avoid, regardless of intentions.

Some other notes:

1. As stated, I don’t bring this up to add to a pile on, just to use this as an example of something I’ve seen happen with folks trying to engage in this tech with an ethical framework. This is because-

2. Frankly, I think shame is a shit motivator. I bring all this up to try to draw attention to something I find under discussed which is what these LLMS are meant to do, and what they are advertised for which are often very separate lists. (I didn’t even get into how they’re meant to train off user data but that’s another can of worms). If you want to discourage AI use, it’s a far better strategy to listen to why the person is engaging with AI and then address that reason.

3. I don’t want to come off as someone trying to scare folks that they’ll get addicted just by touching the stuff: my point is merely that is it critical to know what something is made for before engaging with it. This actually goes for a lot of tech besides AI like Google or social media (which are also designed to 1. Capture your attention and 2. Show you advertisements). AI isn’t a special kind of evil in this way, merely a continuation of a tech trend where a product is sold to a consumer with the goal of the time of the consumer becoming the source of revenue.