Hit and Miss #294: Explaining the inscrutable
It’s an overcast Sunday and there’s a mariachi band playing outside, which, much to T and A’s chagrin, means I’m trumpet-mouthing along (I suppose we should all be glad I haven’t yet broken out my actual trumpet). It’s not been quite the relaxing weekend either of us were expecting (not because of the mariachi band, but because of various continued health adventures), but we’re hanging in nonetheless.
Ottawa’s been thrumming this week with the PSAC strike. For those who wonder why office workers ought to unionize and, yes, strike—though not all PSAC workers are office workers, they make up a significant portion—Anne Helen Petersen offers a compelling argument, grounded in many of her previous pieces on precarity, class, and honest living. A +1 to Sean’s thoughts on the strike.
“AI” (particularly in its current incarnation, large language models, or LLMs, or, more generally, “generative AI”) continues to be in the news. I’ve been looking forward (with great apprehension) to how we’ll fit LLMs and similar technology into current legal frames.
As an example from the last week, the music industry has fought back against an AI-generated song for copyright infringement. Matthew Butterick has written persuasively of the likely intellectual property infringement underpinning GitHub Copilot (and is counsel in litigation against it, along with litigation against Stable Diffusion), and warned of the risks of AI to the rule of law—as currently framed, particularly in the US—where robotic actors are often largely unregulated.
Future directions for this technology will be shaped—though likely not entirely constrained, if (the tech) industry’s history of wilful lawbreaking is any sign—by legal cases to come. And those cases will depend on explaining how these inscrutable technologies work. Tech translation, explaining technology to lawyers and judges, will be essential to how these cases play out—the outcomes depending on one profession’s ability to converse convincingly with another (and, of course, on who the judge is).
This suggests, then, the importance of metaphor and storytelling for the years to come. Winston Hearn has gathered and reflected on various “metaphors for machine learning”, a good example of the kind of hard work that lays ahead. I particularly appreciate how Winston reaches back to thinking and writing from decades past—it’s too easy to get caught up in the excitement (whether deserved or not) of the current moment, overlooking so much good that’s come before.
In a nice change, I’ve had some time for reading in book book land. I highly recommend two:
- Ursula K. Le Guin’s Worlds of Exile and Illusion, collecting Le Guin’s first three novels (!), in which you find the making of her Hainish universe and an already thought-provoking writer (see also Le Guin’s ideal daily routine and her rejection of the idea of “spare” time—what a treasure).
- Tamar Adler’s Everlasting Meal Cookbook, which builds on one of the clever-est chapters in her extraordinary 2011 pictureless cookbook, An Everlasting Meal, “How to Snatch Victory from the Jaws of Defeat”, this time offering dish ideas for odds and ends in the fridge or around the kitchen.
All the best for the week ahead!
Lucas