"Challenging AI's ways of knowing means not only establishing a peer community but questioning what kind of knowledge can actually be created. The insight from standpoint theory is that knowledge is always partial, and not acknowledging this partiality means that knowledge is actually 'taking sides'."
McQuillan, Resisting AI, Chapter 5: Post-Machinic Learning
"The world of experience comes from what Barad, borrowing directly from Judith Butler, calls 'sedimenting out': the way the repetition of specific divisions lays down the more familiar world of 'this' and 'that'. We can understand AI in the same way: as an apparatus whose repetition and recursion sediments out the world that it is supposedly just predicting. It's by interrupting these repetitions that we can reclaim our collective agency."
McQuillan, Resisting AI, Ch 6: Post-Machinic Learning
"Rather than starting from the statistical assumption of separability, which always comes with a gradient of power, mutual aid embraces entanglement as a form of leveling."
McQuillan, Resisting AI, Ch 6: People's Councils
The last chapter talks about the possibilities of workers councils in the tech industry, and in parallel, peoples councils. In distinction to labor unions and their representative democracy, workers councils and peoples councils are bodies of those concerned directly present, their subjectivity directly engaged. This puts them counter to the logic of AI, with its abstraction and denial of subjectivity. Workers councils and peoples councils are based on mutual aid, rather than exploitation.
McQuillan refers to "Tools for Conviviality", which discusses designing tools with limits so they enhance creativity and happiness but do not replace or subordinate human agency and community.
That strikes me as relevant to critiques of FOSS design ethos. What are genuine failings of FOSS design, and what are deliberate limitations intended to foster our agency?
"Tempting as it might be to look at the continent-spanning logistics of a corporation like Walmart and imagine it reordered for social benefit (Rozworski and Philips, 2019], or to wish for accelerated automation across healthcare and manufacturing that delivers luxury for society as a whole (Bastani, 2020), everything we've looked at in terms of the coupling of technical and social orders tells us this won't pan out as we may hope."
(1/2)
"Existing systems, and AI perhaps most of all, are not simply tools that can be turned to good ends or bad but technosocial infrastructures with an established momentum. A new apparatus will bear as much resemblance to existing AI as the eco-social innovations of the Lucas Plan did to the tanks and planes produced by Lucas Aerospace."
McQuillan, Resisting AI, Ch 7: Anti-Fascist AI
(2/2)
"The framing of a new apparatus accepts that the diversity, variety, and complexity of experience overflows representation and is therefore immune to representation."
McQuillan, Resisting AI, Ch 7: Anti-Fascist AI
I strongly recommend Resisting AI, particularly to any radical leftist working in the IT industry.