Friday, 15 February 2019


A research group has trained a neural network that when given a prompt for prose or a news item can extrapolate a whole, plausible sounding passage, which for all its naturalness is wholly a product of machine learning and the computational musings, however, convincing are divorced from reality.
Recognising the potential for propagandise the automation of literature and the press (here’s a less scary example), the group is for now withholding releasing their findings to the public, until such time as we are better equipped to use such abilities in positive ways and not suffer under the abuse of them but I wonder if we’ll ever truly be prepared and sufficiently girded against our own copyediting and desire for a quick turn-over reflected back at us far faster than we could churn it out. Read more at the link up top.