Some brief recommendations (AI, Law, and Otter Things #14)
Hello, everyone! This issue will be mercifully brief, as I will just share some reading highlights. Unfortunately, I could not plan a longer issue this time, but hopefully next week we'll be back to business as usual.
Basically, what happened is that I got a cold last week, shortly after finishing the previous issue. It had been a long time since my last cold, so I was not really expecting to be so unproductive for a few days. This means that now I have to catch up with some writing, leaving no room to develop interesting ideas in this newsletter.
I did not want to simply skip this week, as I am trying to work on my regularity as a writer. Instead, I will present some stuff I've been reading, with some notes on why they might be interesting for you. Of course, feel free to sit this one out and come back next week.
Paul Burgess, ‘Why We Need to Abandon “The Rule of Law”’ (IACL-IADC Blog, 21 September 2021)
When does a concept cease being worth the trouble? This is a question I have explored with regard to the use of “algorithm” in social analyses of computing: the mere fact that computer scientists and social scientists use the term to refer to different things is not enough to invalidate each use. But this ambiguity becomes a problem when it blurs analysis rather than contributing to it. In this blog post, Burgess argues that this is exactly what happens with “rule of law”: the overlap between the various things we discuss under this umbrella is so small that we might be better off referring to things separately rather than mentioning which rule of law one’s speaking of.
Neil C Thompson and others, ‘Deep Learning’s Diminishing Returns’ (2021) IEEE Spectrum
Deep learning approaches are the poster children of modern AI research. This text, however, discusses a factor that might lead them to hit the ceiling soon: the marginal gains in decision quality one might obtain from providing additional data fall quickly. After a certain point, vast volumes of data are necessary for marginal gains in performance, something that is troublesome not just from a data protection perspective but also considering the additional hardware and software resources needed to process this extra data.
Emilio Calvano and others, ‘Protecting Consumers from Collusive Prices Due to AI’ (2020) 370 Science 1040.
A new chapter in the algorithmic collusion debate: in another paper (Calvano et al. 2021), the authors show that collusion between algorithmic agents can be achieved in simulations of imperfect monitoring scenarios. This possibility poses challenges to current anti-collusion approaches, which draws from the communication between actors to identify collusion. Since algorithmic collusion would allow for collusive scenarios with no explicit communication, the authors suggest that detecting it would require mapping the pricing rules learned by the intelligent agents.
Matthijs M Maas, ‘Aligning AI Regulation to Sociotechnical Change’ in Justin Bullock and others (eds), Oxford Handbook on AI Governance (Oxford University Press 2022)
This chapter, available as a (quite advanced) draft on SSRN, argues that AI regulation should not be centred on technological change. Instead, it should focus on the various forms of sociotechnical change that might follow from adopting AI technologies. Such an approach will not only avoid an excessive focus on “flashy” applications but also help us design adequate instruments: should regulation be tech-neutral? How much anticipation is needed? As a result, the paper provides a good framework for navigating sociotechnical change in AI regulation, and its panoramic view of the field also provides a valuable gateway into the debate.
Anupam Chander and Haochen Sun, ‘Sovereignty 2.0’ (Social Science Research Network 2021).
Digital sovereignty is part of tech policy discourses all over the world. While there are sincere reasons for that, and good reasons for pursuing sovereignty, the digital environment opens some unique risks for rights, either due to deliberate action or regulatory side effects.
David J Lynch, ‘Inside America’s Broken Supply Chain’ Washington Post (2 October 2021)
Much of US industry and commerce rely on global supply chains, and any strains to those chains, such as those caused by the pandemic, has compound effects. Yet, there are serious issues with US physical and informational infrastructures that make the American situation worse than, say, what European ports face.
Haris Durrani, ‘Gurney Halleck, the Moor; or, Othello in Space’ (Medium, 7 October 2021)
This essay looks at Gurney Halleck, one of the most interesting secondary characters in Dune. While most cinematic portrayals have coded him as white, Durrani shows that Halleck's physical description and background as a colonial subject are both compatible with reading him as a Moor, and analyses what such a reading implies in the overall Dune narrative.