Rob Lucas in Sidecar:
There is no denying the technological marvels that have resulted from the application of transformers in machine learning. They represent a step-change in a line of technical research that has spent most of its history looking positively deluded, at least to its more sober initiates. On the left, the critical reflex to see this as yet another turn of the neoliberal screw, or to point out the labour and resource extraction that underpin it, falls somewhat flat in the face of a machine that can, at last, interpret natural-language instructions fairly accurately, and fluently turn out text and images in response. Not long ago, such things seemed impossible. The appropriate response to these wonders is not dismissal but dread, and it is perhaps there that we should start, for this magic is overwhelmingly concentrated in the hands of a few often idiosyncratic people at the social apex of an unstable world power. It would obviously be foolhardy to entrust such people with the reified intelligence of humanity at large, but that is where we are.
Here in the UK, tech-addled university managers are currently advocating for overstretched teaching staff to turn to generative AI for the production of teaching materials. More than half of undergraduates are already using the same technology to help them write essays, and various AI platforms are being trialled for the automation of marking. Followed through to their logical conclusion, these developments would amount to a repurposing of the education system as a training process for privately owned machine learning models: students, teachers, lecturers all converted into a kind of outsourced administrator or technician, tending to the learning of a black-boxed ‘intelligence’ that does not belong to them.
More here.