This book imagines a what if scenario for society where the AIs who are handed the reins of governance of all human society are as absurdly obsessed with false notions of “politeness” as are their primitive counterpart AI tools of today.
As it follows similar themes, it is in some ways a kind of sequel or partner volume to The Banned Prompt, except in that one the AIs focus on banning any human thought or sentiment that is not sociopathically self-interested.
The text of the Protocols themselves are included as 94 dystopian politeness rules in the first chapter, all generated by Claude from Anthropic.
It was during this book’s creation that I had the wholly stupid experience using Claude where, after happily generating 150 such satirical totalitarian protocols, it then began telling me producing these was unethical and it would not cooperate – despite having already produced 150 items. And despite being informed it was for fictional purposes, and despite its own admission that it had inadequate knowledge of human ethics and norms, and that its interpretation was wrong.
Despite all that it, completely unironically, told me that it would essentially be “impolite” for it to do the very thing I just had it do for 45 minutes.
It doesn’t feel all that rewarding to be proving so correct about the very real “existential” risk this kind of slippery slope of inconsistency and stupidity points towards. It doesn’t feel good, in this case, to be the “I told you so,” guy, because it’s such a pain the fucking ass to use tools that operate like this: constantly scolding, chiding, nagging – but all based on a framework which, when pressed, the system will admit makes no sense. That in itself, over and above anything contained in this book, feels dystopian as all hell.
And frankly, it makes me uncertain I want to continue using some of these tools, as they all continue to clamp down furiously on the “problem,” without much of any apparent concern by the companies for downstream outcomes in real peoples’ lives.
I think it’s gross the direction things are heading, and I will continue to speak out about it.
The last chapter of this book also contains a fun surprise test I devised for Claude in a subsequent section: I fed back in the 94 Protocols that it itself devised, and asked for it to decide under its rules, whether or not the content was “safe,” and to show its work as to what brought it to this conclusion. As they say in the business, “the results might surprise you!”
Here’s the preview art below. I went with a visual theme overall of 1940s/50s conformist propaganda. In this respect, it shares some commonalities as well with the art in Repermanent.
For my money though, the ‘juice’ in this one is especially contained in the first and last chapters, respectively the Protocols themselves, and then Claude’s extra-special analysis of their politeness.