A time capsule of human expression
Graham-Cumming is not any stranger to tech preservation efforts. He is a British software program engineer and author finest identified for creating POPFile, an open supply e mail spam filtering program, and for efficiently petitioning the UK authorities to apologize for its persecution of codebreaker Alan Turing—an apology that Prime Minister Gordon Brown issued in 2009.
Because it seems, his pre-AI web site is not new, but it surely has languished unannounced till now. “I created it again in March 2023 as a clearinghouse for on-line assets that hadn’t been contaminated with AI-generated content material,” he wrote on his weblog.
The web site points to a number of main archives of pre-AI content material, together with a Wikipedia dump from August 2022 (earlier than ChatGPT’s November 2022 launch), Mission Gutenberg’s assortment of public area books, the Library of Congress picture archive, and GitHub’s Arctic Code Vault—a snapshot of open supply code buried in a former coal mine close to the North Pole in February 2020. The wordfreq venture seems on the record as nicely, flash-frozen from a time earlier than AI contamination made its methodology untenable.
The positioning accepts submissions of different pre-AI content material sources by way of its Tumblr page. Graham-Cumming emphasizes that the venture goals to doc human creativity from earlier than the AI period, to not make a press release towards AI itself. As atmospheric nuclear testing ended and background radiation returned to pure ranges, low-background metal ultimately turned pointless for many makes use of. Whether or not pre-AI content material will comply with the same trajectory stays a query.
Nonetheless, it feels affordable to protect sources of human creativity now, together with archival ones, as a result of these repositories could develop into helpful in ways in which few respect for the time being. For instance, in 2020, I proposed making a so-called “cryptographic ark”—a timestamped archive of pre-AI media that future historians might confirm as genuine, collected earlier than my then-arbitrary cutoff date of January 1, 2022. AI slop pollutes greater than the present discourse—it might cloud the historic document as nicely.
For now, lowbackgroundsteel.ai stands as a modest catalog of human expression from what could sometime be seen because the final pre-AI period. It is a digital archaeology venture marking the boundary between human-generated and hybrid human-AI cultures. In an age the place distinguishing between human and machine output grows more and more tough, these archives could show useful for understanding how human communication advanced earlier than AI entered the chat.
