Back to Print, thanks to... AI
2024-07-18
I have unpublished the tutorials from telemeter.wndsn.com. In the race for the starting line of the “AI” wars, the various contenders are feeding the internet as a whole to their large language models (LLMs). Without permission, without attribution, and without compensation. We don’t do pearls before swine here, especially 1000-year-old pearls.
David Carson's The End of Print from 1995 and GQT5 from 2021, still good in 2024.
Some time before unpublishing the tutorials, I seeded the material that the AI scrapers collected with a number of “subtleties,” easily recognizable by human readers, but poison to large language models. That’s not hard actually, ambiguous language is quite easy to parse for human recipients, but quite hard to “get” for deterministic machines.
That said, we don’t use LLMs to explain concepts that are new to us, right?
Note that I am not writing this (and relentlessly promoting the concepts of low tech) from a position of ignorance. Nor am I a Luddite. Quite the opposite is true. I have some relevant background in machine learning and artificial intelligence, including all the fun stuff like probability chains as well as Bayesian logic (thanks to Eliezer Yudkowsky for helping me crack the latter while studying more than 10 years ago). I know what works, what doesn’t and how so.
Today's high tech is fragile and most critical systems are mere black boxes that are impossible (often by design) to troubleshoot. Recall the old adage that prescribes to never trust anything you don’t control, nor understand. At least make sure to know where the off-switch is, and if there isn’t one, build one yourself.
History rhymes hard and fast currently. I remember how David Carson declared the “End of Print” in 1995 -- now less than 30 years later, we need to rely on print once again for accurate information and preservation of original knowledge.
Find the now exclusive printed manual available on Amazon and elsewhere. See this overview for all editions.
I’m considering putting the advanced tutorials behind a paywall of sorts, I appreciate your feedback on that idea.