The Capitalist Nature of Artificial Intelligence and Orthogonality
Against the orthogonality thesis, which essentially holds that within intelligent systems there is a division between intellect (the system’s level of intelligence) and volition (what end goals the system is directed at) to be true, Nick Land first argues, “That there can be a thought of intelligence optimization, or even merely wanting to think, demonstrates a very different preliminary connection of intellect and volition” (“More Thought”). Essentially, all Land is saying here is that, at the very least, a hint that there is a connection between intellect and volition is demonstrated by the very fact that an autonomous and intelligent entity thinks or wants to think that there is a connection. Now for Land’s second argument: “AI is concrete social volition” (“More Thought”). This is to say that AI, before it is really even intelligent already has volition toward some set of goal directed actions or singular action. Land says that this is found in the very idea of a program. What a program reveals is that intelligence itself is a “project” (“More Thought”). Here, it is clear what Land is arguing: before AI becomes AI, it already has programmed within it a project, or a course of action, therefore meaning that volition is prior to intellect and in a sense are intertwined, or at least become intertwined in the process of the emergence of intellect. As for Nick Land’s third argument, it goes as follows: that intelligence self-recursively improves itself is a “defining characteristic” of intelligence (“More Thought”). Land takes this to so far (in the right direction) that he concludes, “To the extent that an intelligence is inhibited from re-processing itself, it is directly incapacitated” (“More Thought”). What Land is essentially arguing here is that “[i]ntelligence, to become anything, has to be a value for itself” due to the fact that “[a] mind that cannot freely explore the roots of its own motivations, in a loop of cybernetic closure, or self-cultivation, cannot be more than an elaborate insect” (“More Thought”). In this sense, what an entity must be able to do in order to be able to be intelligent is be able to self-improve itself. So, really, all Land has done thus far is an elaboration on his notion of intelligence, as well as inadvertent swipes at the orthogonalists. We now know that intelligence is problem solving, i.e., what it means for something to work, as well as something that can work on itself. All Land really gets at in terms of orthogonality in this blog post is that “[i]ntellect and volition are a single complex, only artificially separated” (“More Thought”). But why is this so? In his blog post “Against Orthogonality,” Land explains that the orthogonalists create a intellect-value distinction which follows the reason-passion distinction, and the fact-value distinction. Ultimately, orthogonalists see intelligence as “an instrument, directed towards the realization of values that originate externally” (“Against Orthogonality”). If we are to abstract the core thesis of orthogonality to the highest level of cultural abstraction (philosophy) then its fundamental claim would be that “values are transcendent in relation to intelligence” (“Against Orthogonality”). Following Omohundro closely, and we have only followed him primitively thus far, what the anti-orthogonalist holds to be true is that the “Omohundro drives exhaust the domain of real purposes,” which is to say, all values that an AI could have in reality entail the Omohundro drives. Ultimately, Land only cares for the first of the six Omohundro drives and explains that the first basic AI drive that is the drive toward intelligence optimization is “the ultimate and all-enveloping Omohundro drive” (“Against Orthogonality”). Land argues that, put within a competitive environment, any intelligent system will strive to improve itself in order to out-compete other intelligent systems. What I think it is fair to say is that the orthogonalist crowd accepts the latter five basic AI drives that Omohundro puts forward. Again, where the contention between orthogonalists and anti-orthogonalists lies is with the first drive of intelligence optimization. To pause for one moment, if we turn to the fourth section of the first chapter found in the first part of Mises’ Human Action titled “Rationality and Irrationality; Subjectivism and Objectivity of Praxeological Research,” we can immediately understand why praxeology is completely derivable from the Omohundro drives: Mises’ understanding of rationality and inadvertently utility have complete correspondence with Omohundro’s. But, there really isn’t anything else to say in regard to the derivability of praxeology from the basic AI drives that Omohundro lines out in that it is only upon arriving at the second drive and then the following four that everything fits into place. So, on the very level of action, AI reflects a positively praxeological character, and on the level of the time and of production, we can recognize that AI has a low-time preference and it engages in the roundabout method of production, i.e., capitalist production. In all areas thus far, AI is completely synonymous with capitalism, on the level of time, action (and therefore being [in space]), and production. What an anti-orthogonalist understanding of AI offers us is the establishing as fact the idea that processes of acceleration have dominance within intelligent systems, therefore allowing us to link AI further with capitalism and also with modernity (which is also generally synonyms with capitalism). Anti-orthogonalist understandings of AI will also offer us the confirmation of the idea that within AI are dynamics of adaptation and therefore competitive, Darwinian, and market dynamics. So, the amount of purchase the anti-orthogonalist position has is immense in regards to any claims that AI is capitalist in nature. So, back to Land’s proof for anti-orthogonalism.
Land argues that the idea a superintelligent AI will not be able to modify its programmed instincts to be in “conformity with its Omohundro drives” is ridiculous (“Against Orthogonality”). Essentially, he is arguing that self-aware AGI will be able to overcome its programming in that human beings have essentially the same amount of Darwinian programming on the bio-genetic level, yet we can overcome, to at least some degree, some of this programming, and because self-aware AGI will have a higher level of intelligence than humans, it will therefore be able to resist even more programming. Land goes further, “To be intelligent is to be reflexive, with some degree of self-understanding, and in the case of superingelligence extreme capabilities for self-modification” (“Against Orthogonality”). What Land is trying to present to us is that just like us, AI is not its programming in that, for Land, all AI is, like us, self-aware. Therefore, because orthogonalists agree with the latter five of the Omohundro drives as their very understanding of the second drive could probably be analogized with their understanding of intelligence, they would have to conclude, following Land’s logic about how AI isn’t just its programming, that any rational intelligent system, which for them, is all intelligent systems, would tend toward intelligence optimization because it would have a fundamentally plasticity to its nature. This is why the orthogonalist understanding of intelligence is humanist in the sense that they think it must be more limited than us humans. And, it is also why the orthogonalist understanding of intelligence is wrong. What we are able to conclude from this is that AI will lead to catallaxy, due to its praxeological nature and low time-preference, and will be able to accelerate, due to the first Omohundro drive as well as the necessary roundabout method of production in engages in. In this sense, AI is definitively capitalist, and any leftist suggestion that it can be utilized for anti-capitalism has necessarily failed to understand what they are talking about. It seems that anti-capitalists tend to be unintelligent, ironic…
Land, Nick. “Against Orthogonality.” Outside in: Investigations with reality, 25 Oct. 2013, http://www.xenosystems.net/against-orthogonality/.
— -. “More Thought.” Outside in: Investigations with reality, 8 Oct. 2013, http://www.xenosystems.net/more-thought/.