Everything You’ve Read About ChatGPT is Wrong
It’s normal for writers to keep repeating sound bites instead of looking deeper
So much nonsense has been written in the guise of outing the flaws of Open AI’s ChatGPT3 that it is hard to know where to begin. Let’s start by saying that this article will not be a systematic evaluation. There are other articles that you can find here and elsewhere that give a rundown of the various features and how they “rate” according to “experts”.
Secondly, the topic here is not the coding capabilities, or the possible uses beyond the primary strength of the chatbot, which is language. The main positive feedback seen is the speed and fluency of the responses. An obligatory mention of how it is “scary” and will improve, and then a jump to whose job is at risk.
Depending on who’s writing (humans I mean) the assessment runs the gamut from, “sorry dude my job is safe” to “we’re all F#cked”. The problem is that the various straw man arguments are based on shallow thought and misguided premises.
The infinite loop of inaccuracy fallacy
For example: the idea that, since ChatGPT 3 was trained using publicly available data, presumably a lot of it on the internet, that a kind of endless loop will form that takes bad data (what we already have on the internet) and then uses it to spit out diluted, worse, bad data, which is then regurgitated and…. You get the picture.
The problem with this theory is that humans are caught fully in this highway to hell already. Dumb humans write things and publish them on the internet, other dummer humans read that stuff and use it to write even dummer stuff, and so on. And then, Ouroboros, you’ve got the situation before ChatGPT arrived on scene.
Furthermore, if you ask ChatGPT to name sources and references you’ll see that, for the most part, they are far better and more carefully chosen than the average human research intern assistant at Buzzfeed. Or Daily Beast for that matter.
Similar to the grammar and spelling, which is, for the most part, impeccable, the sourcing is, if anything, too conservative. Paywall protected research sites come up often, for example.
So, in reality, there could be just about as much chance of ChatGPT “cleaning” and filtering the muck of the internet than there is the reverse, as so many have stated. Mistakes made by a chatbot? Of course. By humans, hahahahahah, to err is human, right?
The Real Situation, a.k.a. you are now a movie producer x3
A better fantasy analogy as to what this newfangled toy is all about would be to think about who pays writers (low level 1st draft writers) and what the output is for those employers when they do.
The obvious example does not even apply. If you want to have a writer churn out “news”, such as for example the inside dope on Kanye West’s new wife, you’re out of luck with ChatGPT, as the end of 2021 is a hard cutoff for available information (as of now).
So, how about fiction then? A caricature of movie making would be a good analogy. Say you have a movie producer, or director, or both, and neither are writers, and they don’t want to buy the rights to a novel or other property to adapt.
What do they do? They come up with a half-baked idea, “concept” in the lingo, and then hire a writer to develop the idea into a first draft of a script. ChatGPT can do something similar.
So what usually happens next? People read the first draft and hate it. They pick it apart and make “notes”. Then, often, they fire the writer who did the hopelessly dreadful first draft, and hire a second writer to do a new draft.
This can go on for months, even years, and all sorts of writers can take a stab at the beast with steely knives until there’s some kind of consensus that it should be made into a film.
Often, this too is not the final version, like when the final scene in the script for “Pretty Woman”, according to the internet bore little resemblance to the film we know : “Instead of Richard Gere’s character’s swoon-worthy climb from his limousine to “rescue” Vivian from her fire escape, Julia explains that the film ended with her being thrown “out of the car, he threw the money on top of her, as memory serves, and just drove away, leaving her in some dirty alley.”
The point? Any form of writing is a creative, infinitely mutable, process. ChatGPT can be a contributor to that process, and can be endlessly “fired” and replaced, currently at a cost of zero dollars.
While that may be no substitute for the glorious Hollywood system described above, the idea that, for some stages of development of many writing projects, it could be a cost effective alternative to hacks and divas, is just about right.
If a continuation of this topic seems appropriate, a more insightful direction for the conversation would be about digital deflation.
It is the amount of output that ChatGPT is able to perform at a very low cost (based on the current price of zero) that represents a continuation of the theme of digital deflation, often a shell game where a “free service” somehow magically turns into a trillion dollar business (which eventually collapses, another story), that will cut deeper into society and the economy.
Please help keep us publishing the content you love
Lynxotic may receive a small commission based on any purchases made by following links from this page
Climate Solutions1 month ago
This Climate Solution is a Sleeping Giant
Books1 month ago
Stress is Part of Life: Burnout doesn’t have to be
AI2 months ago
Adobe FireFly is live and…
Books2 months ago
Trump Indicted by Manhattan Grand Jury
AI1 month ago
Does using ChatGPT Generate Dopamine?
LA Local1 month ago
Superbloom is Back! 2023 Rain has Turned SoCal into a Blooming Explosion
Breaking News2 months ago
Trump is Accused of 34 Felonies