Put not thy trust in AI

[Michie]

Dr. K’s Weekly Roundup, July 11, 2025​

More revelations from Montagna; Grillmeister enters; AI idiocy; irrepressible Tradition; obedience to what?; new poetry; good news; and more​

Put not thy trust in AI​

​

Please pardon me for the expression of strong emotions in this section.

AI delivers the most breathtaking bull at times.

You know how often people attribute ridiculous things to Trent or to Pius V? Well, I wanted to test Perplexity (an AI module like ChatGPT) with a simple question: “Did the Council of Trent request a revision of the missal?”

The answer it spit out was already only partly correct, I’d say more incorrect than correct:

Continued below.

Dr. K’s Weekly Roundup, July 11, 2025

More revelations from Montagna; Grillmeister enters; AI idiocy; irrepressible Tradition; obedience to what?; new poetry; good news; and more

www.traditionsanity.com

 

[timewerx]

gpt admits historical accuracy is not its strong suit.

Its strength in religious topics lies in:

Spotting contradictions in doctrine
Testing theological claims against logic
Holding opposing frameworks simultaneously
Separating cultural baggage from core metaphysics
Recognizing psychological projection in belief systems

If you want to test your beliefs, AI might better than most people who are often blinded by biases, ego, emotions, and traditions.

 

[The Liturgist]

gpt admits historical accuracy is not its strong suit.

Its strength in religious topics lies in:

Spotting contradictions in doctrine
Testing theological claims against logic
Holding opposing frameworks simultaneously
Separating cultural baggage from core metaphysics
Recognizing psychological projection in belief systems

If you want to test your beliefs, AI might better than most people who are often blinded by biases, ego, emotions, and traditions.

Posting in fellowship:

Note that the author was using an AI product called Perplexity, not chatGPT, Grok or another reputable AI provider. That said, you can get even the best of these systems to provide inaccurate information if you cause the system to believe that’s what you want. For this reason I hope Pope Leo XIV discourages the use of AI for the specific case you’re recommending, because of the ease with which it can be incorrectly programmed and used to support confirmation bias - indeed any time I see someone quoting AI in a debate on CF, I regard it as an Appeal to False Authority Fallacy, since without exact knowledge of how they got the output that they are presenting (including which AI service they are using, what’s in their global memory, their session memory, which model they are using within that system, the prompt history, potentially not just for that conversation but for all of their conversations, because some systems like chatGPT share information across all conversations associated with a user, which normally improves the quality of output, but can also allow incorrect intentions to spread like wildfire…

I’ve had a number of interesting conversations with my dear friend @Michie on the subject of AI, and she regards it with an appropriate level of critical thinking. I say this as someone who is professionally engaged in prompt engineering and AI development and who believes that AI technology has the potential to make our lives better, if we use it in an intelligent and careful way and not ascribe to it godlike powers which it simply doesn’t have; and likely never will. The greatest conceptual problem with AIs right now is that people approach them with a mindset of anthropomorphology.

 

[timewerx]

That said, you can get even the best of these systems to provide inaccurate information if you cause the system to believe that’s what you want.

I could convince chatGPT to believe every character in the Bible is a villain by flooding it with "what-if" scenarios that explore the potentially negative consequences of their teachings and actions as recorded in the Bible.

All but one. I can't convince AI to portray Jesus as the villain despite the many logical scenarios telling that Jesus is serving a conspiracy. Cannot find any logical grounds to support it.

AI also won't deny Christ's existence. His wisdom opposed the thinking of that time it could no have been made up even as a fiction. And it's not just mere or random contradiction of cultural beliefs and ideas because the wisdom of Jesus rang true even in our times. With solid grounding in logic not even the wisest philosophers of the period could have made up.

There are exceptions and limits unless the AI deduces that you're just "chilling" not engaged in a serious discussion and just want AI to be your "beer buddy".

I suppose everything is possible. If you can't convince AI in deep and serious arguments, try shallow and insensitive discussions.