Our Anachronistic Algorithms

I’ve written before about culture being stuck in the recursions of algorithms, algorithms as a stultifying bureaucracy, and the strange “flatworld” some are calling today’s Internet run by platforms.

It seems safe to say that every platform is just algorithms in disguise — and algorithms are always looking backwards for data and patterns, then re-amplifying the past and spreading it around again.

Because algorithms feed on the past, time loses its urgency — hours become days, days become weeks, weeks become months, and months become years more often and more mysteriously than ever. Time has become elongated and therefore slippery because algorithms have confused us — what was then keeps popping up as now.

• Sports scores from weeks ago surface unpredictably, leading to momentary disorientation.
• Memes return and recur, signposts across time in what feels like circling the algorithmic drain.
• Older music and movies are rehashed because creators have learned that doing so leads to better results.
  • Because algorithms know how old content performs, they boost things referencing it faster.
  • Reviews of older movies, mentions of older TV shows, and pictures of older actors are privileged by the algorithms.

Algorithms are strangely nostalgic, perhaps because like old fogeys, the past is all they have.

It’s an unexpected twist — the modern Internet has become largely about the past.

Even the world of LLMs is just a new type of algorithm — one that is always looking back because it has no ability to create or realistically predict in any way that goes beyond a burlesque of what we do daily.

• Future results are not based on past performance . . . except in the Algorithm Age.

When OCR became mainstream and journal backfiles were being digitized with alacrity, I once argued that journal archives be put online in separate sites, with a clear demarcation and a moving wall of 35-50 years to limit discoverability (basically for historians) and prevent a past of plausible scientific thoughts from swamping more refined or subsequent findings. After all, given how the past keeps expanding, its ballast is always increasing.

I witnessed the error of treating content of any vintage the same when NEJM published a long-forgotten letter about human growth hormone (HGH) and male virility — a speculative, silly letter published probably for the amusement of the doctors at the time. Ingested by dumb search engines and hungry algorithms looking for clicks, it kicked off the low-T, HGH movement that is now represented by Doug Flutie and Frank Thomas on television.

Years earlier, I has seen this when I was at the American Academy of Pediatrics, as the Wakefield paper kicked of the anti-vax movement. Leadership felt that an article in a journal would fade from memory soon enough. They didn’t count on celebrities, Google, and the click-hungry algorithms of the dawning Algorithm Age turning that one small and obscure kerfuffle into what has become a decades-spanning, commercially-entrenched movement against vaccination specifically and science generally.

The Algorithms care not.

A sense of an algorithm-driven forced march into the past arose again this week, when I picked up Salman Rushdie’s new book, Knife: Meditations After an Attempted Murder, and experienced a sense of being thrown back into the late 1980s, when his novel The Satanic Verses first came out.