Via Web Curios, we are directed to a rather thoroughgoing and detailed dissertation to temper shortcomings and disappointment over artificial intelligence, plus our collective, oblique anxiety over what’s going on under the hood, by acknowledging that the large language models that we collaborate with are ludic (from the Latin for ludus or ludi, games—in the creative and playful sense) technology. The central crux of the thesis by Venkatesh Rao is that we find fault in the onerous tasks we seek to automate by dent of outsourcing what we don’t find fun, unwilling to engage with those missteps and the same impatience manifests when it comes to our expectations, misplaced as the off-registered AI outputs when put to more serious work, in not recognising its toy-like nature.
We’ve identified for the machine a continuum of identity and equivalence for a car, from a sketch, to logo, to image, to video footage, to a model, to an actual car all as the same concept but only the last is not the form of a car with all the inherent physical constraints and requirements, which the AI does not know—being wrong about reality and representation in the same way images and models are, in correspondence with the way we can be duped by deep fakes or confidently wrong answers. Threats to playfulness and toy-making, model-hobbying are instinctive in this context and compensated for through deprogramming, overcoming the perceived infringement on vaunted humanity and genuineness when we are as poised for play and abstraction—which entails taking command of a situation with imagination, marshalling our toy soldiers—and balanced expectations of what we are working with.