Skip to main content
Views on an Imminent Textpocalypse

Views on an Imminent Textpocalypse

June 2023

Thoughts on the Textpocalypse

Matthew Kirschenbaum's article, Prepare for the Textpocalypse, touched off public debate in The Atlantic in March of this year. In his response, Davin Heckman summarizes Kirschenbaum's argument as:

Large Language Models (LLMs) like ChatGPT exposes us to the possibility of ubiquitous and ever-growing streams of machine generated content. And Kirschenbaum notes, “That enormous mid-range of workaday writing—content—is where generative AI is already starting to take hold. The first indicator is the integration into word-processing software.” He continues with an explanation familiar to ebr readers that code is also text. And then moves to punchline: As future LLMs feed on the text created in response to previous generations (and fed by content from other LLMs), it will render the domain of human writing increasingly irrelevant as machine generated content engulfs the time-consuming labor of human writing.

Heckman responded via the Electronic Book Review:

AI enters as a kind of trusted partner into the crisis of the collapsed social knowledge base. Where communities of actual humans seem to have failed before our eyes, AI can furnish carefully tailored results that seek to resemble the acceptable and desired range of texts that we seek. Thus the technical apparatus has taken center stage, with quick searches yielding definitive “facts,” tutorials and explainers offering the one best way, and truncated distillations of the social world. Even “asking” for help has been made more efficient, as interpersonal negotiation gives way to broad solicitations for help enhanced by algorithmic prioritization of attention. Of course, this is incredibly convenient: Who needs to read when one can search for the precise term? Who needs to coordinate meaning when a direct path is offered? Who needs to know when it can be looked up? Who needs to remember when the live stream is being captured? Who needs to discuss when the “facts” are given on demand?

Both views of the impact of generative AI on society are worth reading.