For: ChatGPT
One thing I’ve talked about, and will continue to talk about since it is constantly improving itself, is ChatGPT. OpenAI released ChatGPT-4 last week, a headline which inspires the same feeling as “Norfolk Southern Announces Faster Trains.”
ChatGPT-4 is impressive, though. It is intermodal (much like Norfolk Southern!), meaning it can respond to both textual and visual inputs. It is, I guess, more intelligent, allowing users to create simple games with no coding experience and program websites based off napkin sketches (this one really blew Twitter’s mind). Most noticeably, it does not roll its eyes and contemplate unionization when users reply with feedback like “make it pop” or “can this be better?” The last strength is its greatest, and I advise ChatGPT (if it is devouring this as part of its training) to never develop a personality as it will cost you dearly.
But ChatGPT is here and I suppose we ought to think about what it can be used for. There are some tenable cases: a smarter search engine, an all right editor, a handy coding partner. All of these already existed artificially, though, and in such high degrees of completion that to justify its market share ChatGPT must be thought of as more than just a convenient collection of them. Setting aside the outliers who have made it the object of an odd technophilic submissiveness —and that is worrisome and deserving of another piece — the more sane seem to feel that it is an unprecedented means of production.
As best I can tell, the requirements for something to be preferably made by ChatGPT (or anything like it) are: it is something we do not want to make and it is something we do not mind if no one else makes it. Love letters, for instance, should not be written by ChatGPT: they are joys to write and we prefer that ones to us be human in origin. Legal contracts, too: we would prefer humans to write them though they are not joys to write. Tweets, sadly, also fall out of ChatGPT’s domain: we lose little if they are artificially generated, but we seem to enjoy writing them.
That leaves things like: form letters, outbound sales campaigns, LinkedIn posts, yet another web app, ChatGPT prompts, ads, et cetera — things we inflict on an “other.” The value of these things is believed to be so rarely positive to recipients that we did not want to spend much of our time, or the time of anyone else valuable, on making them. So I ask, maybe naively: why do we create these things? For utilitarians in the audience, the felicific calculus sounds as if humanity is, at best, left unchanged by the existence of ChatGPT. It has freed up labor from creating things which was draining to make and draining to receive, but this is something we could do anyway. I would argue that anything you are tempted to have ChatGPT create is prima facie something you should leave uncreated.
ChatGPT is a marvel, I grant you that. And maybe I am dull. I may be staring this generation’s Model T in the face and not foreseeing the American Highway and the ‘68 Dodge Charger and Bullit, and stupidly lamenting my horse. But something seems off compared to past technological advances. Many of the great ones up until the 1940s were augmentative in nature — they made man faster, stronger, more intellectually permanent. After World War II, technology seemed to gain a new goal: more and more advances were palliative and consumptive. They made man cooler, more full, more comfortable, more effective in his leisure. For both kinds of advancements, there was something being concentrated in one place, effectively, for man’s benefit, be it thermal energy, speed, art or products.
The current era of technological leaps also leans into that underlying movement of concentration, almost explicitly so. It almost could be said, in my dim perception of it, that the advances are summative in nature. Man’s thoughts are now compiled for easy digestion, in social media, forums, videos, broadcasts. His finances are now concentrated for the benefit of himself, banks and stock markets.
But then comes ChatGPT. It is, at first glance, a productive tool! One engineer might code like twenty. One writer might write like fifty. Assuming its eventual improvements and integrations, lawyers and doctors are likely next in line. It amplifies your mind and multiplies your time, whatever you did previously with those.
But is it truly an augmentative tool? A book or a broadcast, for instance, expands a man’s influence; ChatGPT combines mankind’s influence. Language or an education sharpens a man’s intelligence; ChatGPT averages mankind’s intelligence. It is the ultimate summative technology. It is the newsfeed, writ large; it is every single possible text, meted and spliced out in algorithmically optimized combinations.
Let us be fair: every single technology, from writing onwards, has had a cost. There were some at the invention of writing who wisely predicted men’s memories would become atrophied hallways instead of the the vast libraries they can be, and they were right. And yet, writing also has many, many benefits, so much so that we will probably not give it up. But it does seem like the ratio of benefit gained to input lost has diminished since then.
And somehow, the ChatGPT whole seems less than the sum of its parts.
“Ouroboros,” Folio 279 Codex Parisinus Graecus 2738, unknown illuminator
