Week 7 - Temporal aspects of big data/AI - Core Post 4 (Hyejoo)

One thing in particular from Wendy Chun’s chapter that gave me a really new perspective was the idea that big data hype, so dependent on correlations, is all about reconstructing and perpetuating the past––it betrays a regressive temporal desire: “Correlation’s eugenicist history matters… because when correlation works, it does so by making the present and future coincide with a highly curated past” (52). I’ve never thought about a temporal aspect of AI, in the way that all of AI’s predictions and future-oriented learning rely on a preexisting set of correlated data. Big data’s futurism is ironically a past-ism. I am more convinced, therefore, of the importance of “imagination,” as we’ve seen already in Kara Keeling and Audrey Anable’s pieces, of creative thinking that envisions designs and systems that really break apart the correlations of what we already know or perceive. Indeed, Chun picks up on the actual lack of imagination in big data: “The important point is that predictions based on correlations seek to make true disruption impossible, which is perhaps why they are so disruptive” (52). Imagination is limited if everything must be modeled after the past. Though, I think the question remains of whether or not all AI is by definition based on correlations, as AI is inherently based on data and algorithms. Thus, does AI by definition inescapably rely on the past? Can AI ever truly escape the past-ism of correlations?

Crawford and Joler’s essay also mentions the past-ism of AI systems, for they “repeat the most stereotypical and restricted social patterns, re-inscribing a normative vision of the human past and projecting it into the human future.” We’ve seen this, too, in Noble’s research on Google searches. But Crawford and Joler’s emphasis on the tangible materiality of tech supply chains also shows how products like Alexa, just to power its machine learning, literally destroy the past/present (e.g. the extinction of Malaysian latex trees or the exploitation of cobalt excavation workers) while inherently relying on it. The destruction is even more upsetting as, again, all of it is for lofty notions of progress in technology, when so much of it actually just perpetuates a reified model of the past. The temporal irony is almost absurd.

I was not as convinced by Cheney-Lippold’s fear of “measurable types.” Perhaps I got a bit lost in his nuances around “protocategorical intersectionality” and “posteriori politics” (84–85), but nonetheless it seems C-L is mainly problematizing how data categorization, collection and analysis do not reflect our lived experiences by disregarding our a priori histories and contexts. “It’s missing the referent by which one can know, and live, one’s identity” (85), and so the computer talks “about me, not to me” (91). Yet, I’m rather glad the computer does not actually talk to me. Maybe C-L gives a little too much power to the data collectors in shaping the way we view ourselves. Not that this isn’t unimportant or uninteresting. After Chun and Crawford and Joler I’m also left wondering how the correlative, extractive logics behind data-driven digital tech also change and reflect how we understand our subjectivities. But in the case of C-L’s critiques––do I actually want computers to know and accurately reflect my lived experience? Not really.