The problem with AI
By Chris Edwards and ChatGPT-4
ChatGPT’s latest iteration is an advanced language model developed by OpenAI that employs the GPT-3.5 architecture. This iteration has been trained on a vast amount of data to improve its natural language processing abilities, making it capable of understanding and generating more human-like responses. Additionally, it has been optimized for speed, allowing it to process and respond to queries at an unprecedented rate.
The latest version of ChatGPT incorporates state-of-the-art technologies such as self-supervised learning, meta-learning, and transfer learning. These techniques have enabled the model to learn from vast amounts of unstructured data, resulting in a more accurate and contextually appropriate understanding of natural language. As a result, ChatGPT can answer complex questions, generate human-like responses, and even perform language-related tasks such as summarization and translation with a high degree of accuracy.
So, gentle readers, a bit of disclosure to frame this outing: the above two paragraphs were written by ChatGPT’s latest version. I simply entered the prompt “Give two paragraphs that explain ChatGPT’s latest iteration in 70 words or less,” and off to work it went.
Now this column might come as something a bit removed from my jurisdiction, but I’m writing it anyway. By the way, before I get into my rant, good-n-proper, there’s an irony present that, yes, last week I wrote about how awesome Artificial Intelligence (AI) was because it could make all the presidents have mullets and mirrored motorcycle cop shades, but I digress.
Typically, things of a futurist bent, or anything having to do with technology and/or computerized legerdemain is the purview of our Jim-of-all-trades and occasional op-ed page cohort Jim Powers, and I’m sure he could posit a few thoughts on this subject in an even more compelling and cogent fashion, but here goes.
AI-generated writing is already fairly commonplace with certain types of writing. Many businesses and organizations utilize it to create financial reports and even copy for marketing.
The vast amount of data that the current iteration of ChatGPT uses to create prose from user-supplied prompting draws from all manner of sources – from e-books and Wikipedia to blog entries, yet generates something new. This presents a huge problem in the academic world, as anything from freshman composition essays to graduate-level theses can be generated through AI, and standard plagiarism detection methods are futile.
There are scads of users on Reddit who are reporting having used ChatGPT to get near-perfect grades in college classes, and what would traditionally have taken a student many hours, spread over days, of research, reading, drafting and revising can now take 20 minutes. An AI-generated doctoral thesis that made the rounds in the news cycle even included references, citations and annotations.
AI can write well, to be certain, but it lacks the self-awareness and nuance that comes about through stylistic consistency, building upon strengths and perspective of a human writer. This is especially important in endeavors, such as novel and short-story writing and screeds that fill newspaper op-ed pages.
But what about the casual reader who doesn’t really care about certain craft elements? For those readers who think Dan Brown and Fifty Shades of Gray is good writing, AI-generated writing would likely be good enough to fill their literary needs.
One story I cut from my short fiction collection that was released last year is a piece called “Equanimity.” It is a brief character study about a woman working in corrections and counting the days until she is retired from TDCJ, simply so she can enjoy smoking marijuana.
Lately I’ve been re-writing the story a bit to use in another project, and since a draft was on my desk when I began adventuring in AI writing, I decided to feed the plot into Chat GPT-4 and what it spat out was a nearly 600-word piece that was eerily similar to my original story, and even featured a scene with the character’s retirement party described.
So how long is it before AI writing becomes good enough to completely replace human scribes altogether? Our computing speed continues to accelerate, and I’m told the amount of power available to AI is expected to increase 1,000 times over the course of the next five years.
As someone on the margins of the creative community, I’ll say this here and now: AI is a problem. However, for those of us who do bang words out, whether it’s for a living or our own edification/practice, let’s keep writing and honing our craft. The AI might win someday, but we ain’t going down without a helluva fight.
You are a guest
or post as a guest
Be the first to comment.