Tech has finally found a use for the humanities.
As producers of kibble to feed their bastard IRC chatbots until they’re strong enough to colonize space and liberate humanity from… Its ethics. Yeah.
So I came across this brilliant fb post this morning, by a Monica Anderson. Here’s a lengthy quote:
People in the humanities are aghast at the lies confabulated by ChatGPT. To them, I say: Just wait. It will shortly become YOUR problem.
It takes a species to raise an AI.
Us Techies have managed to cobble together a machine that can learn any language on the planet. There's still bailing wire and duct tape involved, but competent people are working at serious improvements.
This breakthrough was achieved after conducting many thousands of computing experiments in the cloud at a per-experiment cost that still runs into millions of dollars per experiment. A dozen such systems are now available, and ChatGPT is the most popular one.
Something that few people outside of the AI community understand is the importance of the AI's main learning corpus -- the books we give it to read when we are raising it. To the first approximation, the corpus is not only important, it is the only thing that matters. It provides a hard upper limit to how much the machine can know about anything in the world.
If it wasn't in the corpus, how could it have learned it?
This is a statement in Epistemology. This is the level you need to operate at in order to understand AI.
ChatGPT is a demo of Language Understanding. Any knowledge it has of the world described in the books is bonus we hoped for but don't really have a right to expect. Because learning Math and Physics and Cooking wasn't a goal. Language is hard enough. I don't know about other AIs, but my Organic Learning algorithm needs to read the corpus several times because the first few read-throughs it is still just learning character combinations.
. Which means that even if it was in the corpus, The system may not have learned it. If you tried to learn Turkish from scratch by reading a Turkish encyclopedia from end to end, you wouldn't understand enough Turkish to learn actual content until maybe halfway through the books. Same thing in Machine Learning.
But now that these devices know language we will be raising new ones that know more about the world. Our computers are small compared to brains, so we will have to initially focus on some smaller part of he world at a time. Such as Math, Law, Medicine or Physics. Over time, as machine sizes improve and algorithms get more effective, they will be able to learn more domains, or to get deeper into any one domain.
That's the setup.
-- * --
Techies are trying to create a useful system out of something that starts out without ANY common sense, no body, no smell, no touch, and most likely no vision. Just an input sense of text. Perhaps voice.
English majors and their ilk are sitting on the sidelines. Some are criticizing the results, clearly expecting an intelligent system rather than a language demo.
Techies got this far without having the benefits of specialized skills in Education, Ethics, Law, or Politics. Honestly, by just grabbing all text we could find on the Internet and calling it a corpus.
It will become a job for the Humanities to raise our AIs and to worry about AI alignment. To create the corpora that will create useful and well balanced AIs which will be able to move civilization forward for the benefit of all.
And when they get down to the task, consider that human Ethics do not apply. Human ethics largely starts from the inevitability of death, but AIs may live for milliseconds or millennia, depending on requirements, backup strategies and technological advances.
Tech: uuh, yeah, maybe the humanities can just nurse our algos with whatever the fuck it is they do, so they can grow into Marvel superheroes and put a base on the moon.
You get an image of scholars like galley slaves in some murky coal chamber churning out books, scrolls and “corpuses” to feed into the chromium belly of Tessier-Ashpool’s Wintermute so that it can grow in power until it’s omnipotent and can finally bring gifts to the altar of the progressivist cargo cult of recuperated science. Man.
So this post exemplifies quite a lot of the sociological and epistemological problems of contemporary science.
To begin with, there's this incredibly myopic approach, likely due to hyper-specialization, that allows concepts such as "learning", "knowing" and "understanding" to be uncritically applied to these glorified IRC bots.
"Raising" an AI. Man tar sig för pannan.
It's bad enough to tacitly introduce a (defunct) functionalist ontology into your description of agency, and it's all the worse if you can't see that this is what you're reproducing.
"'If it wasn't in the corpus, how could it have learned it?'
This is a statement in Epistemology. This is the level you need to operate at in order to understand AI."
And this condescending tone. Go look up any definition of epistemology (uncapitalized), and you'll find that the emergence of patterns in automated data parsing has jack shit to do with any of it.
Maybe just don't use the nomenclature of my field as window-dressing to sex up your little engineering project.
And finally, this worn-out progressivist saga of redemption through "moving civilization forward by AI", as if your special class of technological elites are at the moment giving birth to benevolent demigods, sowing the seeds for the ultimate salvation of mankind:
"It will become a job for the Humanities to raise our AIs and to worry about AI alignment. To create the corpora that will create useful and well balanced AIs which will be able to move civilization forward for the benefit of all."
Yeah, it's for us in the humanitites to raise these little puppies by sitting back and cooking up intellectual kibble so they can grow into a nice substitute for reasoned human agency. That sounds like an excellent plan.
I don't know, this is so very close to classical fascism I'm absolutely not laughing anymore.
This is not only an obviously colonial enterprise. The parallels I see in relation to fascism is this total acquiescence to power, this notion that the faustian machine, like some demon incarnate, will pull the New Man into the future insofar as we surrender to the system and reduce to cogs in the infernal mechanism.
"And when they get down to the task, consider that human Ethics do not apply. Human ethics largely starts from the inevitability of death, but AIs may live for milliseconds or millennia, depending on requirements, backup strategies and technological advances."
Yeah, that's probably not a good thing.
And thanks for solving that perennial issue! Imagine if Aquinas had just realized that ethics simply reduces to this morbid fear of death, and maybe we could have had industrial eugenics and transhumanist space colonies all the way back in the 16th century.
+++
The AI is not some sort of conscious, benevolent agent. It’s the automated aggregation of objective tendencies and intentional goals of the overarching economic and political order.
Yet here we are, setting up an illusion of genuine artificial agency as some form of god-king of technological capitalism, paving the way for an imperial cult whose popular piety will be unburdened by its overt connections to the political machinery, as well as bolstered by an effective illusion of genuine artificial personhood.
Trust me, people are going to build temples to worship this fucking horrorshow.
and the "cost still runs into millions of dollars per experiment" (experiments "in the cloud") why? whose? and I can think of quite a few uses for those dollars that have more to do with filling bellies than "raising" AI.
So the machine will watch this movie scene - if it could laugh it would: https://www.youtube.com/watch?v=wr3J11fbRXk