Harry Gumboot, Nimbin GoodTimes
It was ANZAC Day when the algorithm suggested I watch an AI generated documentary on Australian troops in Vietnam.
It concerned the way Diggers had a far better success rate in the jungle than the Yanks*, due to factors that included moving slowly, using hand signals rather than speech, and failing to use aftershave (sic). Pretty basic stuff if you want to avoid alerting the enemy.
The thing that irked me was the diggers appeared to be wearing WWI vintage French helmets. AI has trouble with helmets, especially the subtle differences between Russian, German and American models.
In this case they retained their distinctive central ridge, despite it having been replaced during WWII, around 20 years before Australia became involved in Vietnam. However, considering France’s colonial history in SE Asia, this mistake provided an insight into AI’s propensity to hoover up everything before trying to make sense of it.
AI was also having trouble with the Aussie slouch hat; relentlessly showing the turned-up side regardless of which direction its wearer was facing.
I dug deeper and found a small army of similar docos: each the result of AI repeatedly plagiarising former versions of the piece after having sacrificed its human source to the electronic fog of war.
The AI-generated voiceover, accompanied by annoying subtitles, replete with bad punctuation, absent phrase boundaries, mis-matched homonyms, and arbitrary pauses, left things as clear as the aftermath of a well-aimed smoke grenade.
Separating fact from fiction was akin to navigating the labyrinthine layout of a Viet Cong tunnel system. What should have been a five-minute work ran closer to 30. Anyone with a reasonable grasp of English could have edited it down to the former in less than an hour.
In an unrelated video – on a site I chose to subscribe to, rather than have the logarithm suggest – a real scientist was describing how cognitive artefacts – tools we use for thinking – can be divided into complementary and competitive subsets.
Complementary objects, such as pencils, notebooks or maps help to make us smarter; competitive ones, such as calculators, predictive text or GPS, tend to make us dumber.
It’s a metaphor for the point AI was struggling to make: that Aussies tended to do things differently to the Yanks* and, as a result, achieved different outcomes.
*I have avoided the use of the term ‘septics’ lest the algorithm bombard me with posts devoted to sewerage systems.
This article appeared in Nimbin GoodTimes, May 2026.
