Read on Twitter
I'm going to do something very ill-advised and elevate a subtweet to an object-tweet. Let's actually take a look at one of the steaming hot takes that @dynamic_proxy and I have in mind here, David Golumbia's 'The Great White Robot God' - https://davidgolumbia.medium.com/the-great-white-robot-god-bea8e23943da https://twitter.com/dynamic_proxy/status/1342264602982313984
What's wrong with this piece? Where to start. Let me begin by saying that there's not *nothing* here. There are various factual claims, and even the occasional generalisation with a kernel of truth hidden in it, but otherwise the piece has big yarnwork energy:
Everything is framed in terms of vague 'connections':
"To someone writing from my position, it is absolutely true that nearly everything in our society is connected to white supremacy. At this level it is trivially true that AI in general is connected to white supremacy."
"To someone writing from my position, it is absolutely true that nearly everything in our society is connected to white supremacy. At this level it is trivially true that AI in general is connected to white supremacy."
But what kind of connections? Symbolic ones? Social ones? Causal ones? The notion is semantically flattened in such a way that one can slide from one register to another like a greased otter. Symbols become social cliques become causal mechanisms driving us toward *bad politics*.
And it's that *bad politics* from which the rhetorical urgency of the piece derives. This urgency is what makes the greasy slide from one point to another seem like an argument, with inferences, rather than mere associations.
I've complained about 'critique' a lot recently. This particular style of 'critique' has spread like a rash across the humanities. I personally think this is because of the role that literary theory played in diffusing certain philosophical ideas across the humanities as a whole.
This technique of guided association is completely appropriate to literary interpretation, in which there is a direct connection between symbolic content, narrative causality, and the social context within which the one makes sense of the other. This can produce nuanced readings.
But it's not a good way to interpret reality. It makes no sense to paint one's political opponents, no matter how heinous, as either protagonists or antagonists in the causal chains of events you're supposedly exploring, or to interpret the content of their ideas symbolically.
I'm not even saying you shouldn't psychoanalyse your opponents, because this is obviously fruitful in some cases. There's space for the hermeneutics of suspicion, just not for the suspicious metaphysics implied by its methodological overextension. 'Connection' is not enough.
But, if I'm going to critique Golumbia, I need to avoid falling into the type of discourse I'm criticising, so it's important to distil the specific claims he's making and say something concrete about the kernels of truth within them, and the fetid falsities they're wrapped in.
The piece is putatively about the notion of Artificial General Intelligence (AGI). Golumbia gets the basic facts right about the emergence of the term and what it indexes in the field of computer science itself. He's even right about its significance in some social networks.
I've got nothing against the sociological observations regarding LessWrong, online 'rationalism', 'new atheism', and even NRx, Molbug, and Land. It's not exactly groundbreaking, as this is pretty easy to glean. But it's not exactly wrong, even if it isn't interesting.
But the big claim that Golumbia wants to make is that AGI is not so much a concept as an ideological cypher. The fact that there isn't a consensus definition is transformed into the suggestion that, once one unmasks this spook, one reveals the face of white supremacy beneath.
If you think I'm exaggerating the content of Golumbia's claims, or misunderstanding its rhetorical subtleties, please just LOOK at the image accompanying the piece:
So, what reasons does Golumbia have for thinking this? Let's start with the reasons which, by his own admission, he doesn't have. Consider the following quote, in which he addresses the computational theory of mind that he takes to motivate the notion of AGI:
"The brain, this story goes, is a computer […] or at least very similar to a computer; what happens in the mind is what happens in the brain, and what happens in the brain is largely either “information processing” or “pattern recognition” or both...
...(note that I am not specifying the meaning of these terms with anything like the formal precision that would be necessary to interrogate these claims; neither, I am suggesting, do AGI proponents)."
WTF, right? This boils down to: "I don't need to actually talk about the details, because there are no details to talk about." How fucking convenient for you David. You get to bracket out every detail relevant to the discussion of AGI, except your own yarnwork connections.
I'm not claiming that *suspending* a live theoretical debate like this in order to study its surrounding context is impossible, or even undesirable. But as I've said before, one has to allow it to *resume* afterwards. Golumbia's attempted unmasking is intended to prohibit this.
"Don't worry if you don't know *anything* about active debates in computer science and philosophy of computer science, I don't know anything about these things either, and I can still *refute* this idea for you!"
So, what arguments does he actually present then? There are basically three: i) general intelligence qua IQ is systemically racist, ii) computationalism is disembodied, and even if that's not racist (it is) it's bad, and iii) the people interested in AGI are bad (and racist).
Here's my quick response: i) there's no conceptual link between AGI and IQ even if there's a symbolic one, ii) the embodiment paradigm has its own flaws (normative as well as explanatory), and iii) some of them, yeah... but so what?
I forgot to mention that there's a fourth argument that cuts across these registers, and ties the piece together, evident in the title: iv) AGI is implicitly theological, and we can detect within it an inchoate (and sometimes explicit) desire to immanentize the (racist) eschaton.
This is in some sense the most interesting point, but it's also the most lazily articulated, and seems to be there only to function as a framing device that ties the rest of the yarnwork together: Pepe Silvia = Roko's Basilisk = The Great White God
I know my rejection of these arguments is very terse, but how is a leftist who works on AGI like myself supposed to respond to such a thinly veiled accusation of spiritual corruption, which doesn't even pretend to engage the actual topic of my work?
Arguments made in such spectacular bad faith don't invite or deserve responses made in good faith. I've thought about writing a more formal rebuttal to this piece before, but there is so little substance to it that such a piece would mainly be an exposition of my own ideas.
And in that regard, I'm better off just directing you to the places where I've engaged the subjects relevant to Golumbia's claims in a way that makes the substance of my rejection plain, without bending its form to fit what amounts to competently formatted yarnwork ravings.
(i) If you want to know my thoughts about the problem of defining AGI, the best place to go is 'Autonomy and Automation', which has the added advantage of being framed within an overarching discussion of normative political issues: https://deontologistics.wordpress.com/2017/12/22/autonomy-and-automation/
I don't specifically discuss the issue of race there, though I have mentioned it a few times, and have an upcoming post on the topic. However, this post lays out some methodological ideas about systemic failures that include racial bias: https://deontologistics.wordpress.com/2019/11/04/tfe-incompetence-malice-and-evil/
(ii) If you want to read my charitable take on the embodiment paradigm and its flaws, look at my recent interview ( https://www.thephilosopher1923.org/interview-wolfendale). If you want a more vehement critique, there's a blog post that should drop soon.
(iii) If you want my take on NRx and how I think it and positions like it should be resisted, there's my old piece on accelerationism ( https://deontologistics.wordpress.com/2018/02/18/ofta-so-accelerationism-whats-all-that-about/). Hopefully I'll write more about this at some point.
(iv) If you want my take on the implicit theology inherent in those distinctly terrible strands of thought about AGI associated with LessWrong and other so called 'rationalists', then this is the piece for you: https://deontologistics.wordpress.com/2019/10/22/tfe-immanentizing-the-eschaton/
Finally, if you're wondering why I haven't done the noble thing and @'d Golumbia directly, turning a subtweet into the twitter equivalent of a thrown gauntlet, it seems that Golumbia had already preemptively blocked me before this thread even began. Make of this what you will.
