For those of you we’ve welcomed recently to Heir, allow me to introduce one of the blog’s prematurely venerable traditions — the Digestive Post. The Digestives are glorified hot-takes, essentially, drawn from recent life and reading and occasionally even culled from my responses to other Substack writers’ output, intended to clear the gut of the mind of any half-broken-down/assimilated material and, yes, perhaps make way for the heir to the thought.
Today, something that ought to make you feel old — a commemoration of a much beloved classic novel on its anniversary, and an examination of how people do their laundry these days and, particularly, what bits of it they elect to leave prominent on the line.
This is from the OpenAI post on the lawsuit recently filed against them by Elon Musk.
While it was hoped by some commentators that the filing would act as the starting pistol on a controversy as engrossing as the aborted Altman ouster, I cannot say that I find it surprising that it hasn’t been so, and moreover that the public at large are not that interested in what’s going on at OpenAI. I believe, in fact, that the lack of interest in these proceedings is inherently tied to a facet of OpenAI and their output that I am about to discuss.
Consider the list of accomplishments that OpenAI enumerate above as a sort of apologia for their mission. With the exception of the second item, it struck me as a considerably less impressive laundry list of accomplishments than one might have have expected the company to be able to put out at this stage. ‘Marginal gains against the unstoppable tide of bureaucratic tyranny’ is a fine trophy, but the laundry cited is distinctly short of “OpenAI: Your world. Turned upside down. Today.” Which I believe is what we were promised.
The items on the list are not unimpressive in absolute terms — especially not 100xing farmer income in India, which is so much bigger than the other citations it makes them (and particularly the item in front of it) look ridiculous. Nevertheless, these citations simply do not evince the sort of scale impact OpenAI would want to in these instances, nor do they reflect what I would imagine are OpenAI’s priorities as a business entity.
If you made a crude model of what you would generally expect those priorities to be, you can quickly draw from them an appropriate thesis to encapsulate the company’s achievement so far:
“Catalysed the development of an entire industry sector expected to be worth $1t within the next seven years”
You could imagine a dedicated OpenAI milestones page declaiming something like that.
“Reduced time-to-knowledge access for x people by y”
You could imagine this kind of schema being used as well.
But the notion that the company’s leadership would shortlist chits and then think “helping a minor state enter a trade bloc that hates companies like OpenAI” should be the one they lead with just seems daffy. It may, as an associate of mine suggested, be a particularly bald-faced attempt to curry favour with EU legislators, but that in and of itself strikes me as a concession to the actual stakes. OpenAI only recently released to market one of the most shocking products that said markets had ever seen. And yet, the scale of the presence which that product actually commands in the world has not, thus far, been commensurate to the shock. If it did, and if its commercial utility was so irresistible in its present form, no such trade bloc ingratiation would be required. If anything, the flow of ingratiation might be flowing the other way.
You do not, in short, notice ChatGPT’s presence in the world. That is, until you do.
And here is the part of the laundry that no legacy media operator is interested in, and which OpenAI themselves would be loathe to have highlighted, so contemptuously does it undercut the loftiness of their mission and so utterly does it reveal that they are, for now, in league with the most pollutant entities of the net’s intellectual life.
That’s to say: at present measurement, the main legacy of OpenAI is, in the main, the same legacy of all of the major entities in the history of the consumer internet — waste. Waste of infinite exotic varieties, and produced in astonishing volume.
There is, first and foremost, the waste produced in what few would disagree is the generative AI’s natural productive environment — the repo. Research has shown that ChatGPT has exerted considerable downward pressure on code quality since it became a widely assimilated part of the toolstack. From Gitclear:
“We find disconcerting trends for maintainability. Code churn -- the percentage of lines that are reverted or updated less than two weeks after being authored -- is projected to double in 2024 compared to its 2021, pre-AI baseline. We further find that the percentage of "added code" and "copy/pasted code" is increasing in proportion to “updated,” “deleted,” and “moved” code. In this regard, code generated during 2023 more resembles an itinerant contributor, prone to violate the DRY-ness of the repos visited.”
A synchronised increase in code volume and decrease in code quality is one thing. As Erik Hoel recently showed in an awe-inspiring compilation editorial, there are the SEO heists, grinding down the knowledge estate for a cheap, brief, and trivial search traffic kick…
…there are the non-human AI-generated authors employed by mainstream publications and positioned as real people, churning out sub-human-standard AI bilge copy…
…and perhaps most memorably of all, the children’s entertainment videos. If you thought kids brought up by iPads were already bound for negative development outcomes, one can only make distant fathom of what children brought up on these monstrosities will turn out like. Let it be known, for those without the stomach to sit through the video linked above, that it and its ilk are defined by:
Storylines that are openly incoherent, instead of merely facile as most early-years entertainment is pitched at being
Scripts that abound in bad grammar and non-sequiturs
Pseudo-educational tasks that clearly have no particular educational logic
Imagery that is abidingly cheap and devoid of any of the grace or light that even babies can identify preference for in art
Of all the products of generative AI, these are undoubtedly some of the most embarrassing. It is impossible to tell whether AI-waste will, in weight and influence, prove more practically influential across any given timescale than the technology’s more high-minded deployments. Still, it cannot but be said that one of the primary industrial lanes of this technology is the production of pure rubbish, and with it an amplification of the grift for which the net had already proved such a nauseatingly terrific instrument.
It’s hard to imagine the corpus of Drew Ortiz appearing on OpenAI’s highlights page any time soon. Not least bearing in mind the media proper’s complicity in such things, it will be kept absent of wider public attention. It seems, as many before me have pontificated, the number one rule of the international competitive AI fight club is that you do not talk about what your AI can do, or has done. For all those reasons mentioned above it would be a flattering maxim.
Speaking of which.
“You Do (Not) Talk (All the Time) About Fight Club”
I recently saw a retrospective review for David Fincher’s adaptation of Fight Club, on the occasion of its 25th anniversary.
Keep reading with a 7-day free trial
Subscribe to Heir to the Thought to keep reading this post and get 7 days of free access to the full post archives.