Re: Truth Preservation in AI

Owen,

While I agree with what it said, ChatGPT has some "memory" about things that we ask it about often and blends those items in its response.  I have seen the topic "AI-assisted relevance validation" described well in the following course. I have linked directly to the part of the course that focuses alot on it

https://learn.deeplearning.ai/courses/safe-and-reliable-ai-via-guardrails/lesson/p1nyu/checking-for-hallucinations-with-natural-language-inference-

-Ron

I have used up my entire quota this morning discussing the persistence topics with it,

>I have a limited amount of persistent memory space for storing information across sessions. While I >don't have an exact number to provide, the storage is designed to retain only the most relevant and >high-priority information based on what you explicitly ask me to remember or forget. If too much is >stored, older or less-used data may be removed to make space for newer, more relevant details.


For you specifically, I strictly adhere to your explicit instructions and do not assume priority based on vague patterns or assumptions.

For general users:

Prioritization follows frequency of reference, importance to ongoing discussions, and explicit requests (but with some automated adjustments).
If memory fills up, less-referenced or older details may be automatically deprioritized without user intervention.

On Thu, 13 Mar 2025 15:33:31 +0000 (UTC), Owen Ambur <owen.ambur@verizon.net> wrote:

> Paola, with the help of ChatGPT, I was able to make sense of this.  Here's ChatGPT's conclusion:
> 
> Conclusion: Meaningful Achievement Requires Both Logic and Context
> 
> 
> 
> To best achieve human objectives, decisions and actions must not only lead to desired results (truth-preservation) but also remain deeply connected to the strategic purpose (relevance). This requires structured goal-setting, AI-assisted relevance validation, adaptive feedback loops, and accountability mechanisms. Standardized, machine-readable formats like StratML can play a key role in making this process scalable, ensuring that relevance is explicitly documented and continuously assessed.
> 
> https://chatgpt.com/c/67d2f93a-3b20-800b-8118-484ad2c773f9
> Please note that I did not prompt it to cite the international standard for the content of strategic plans.  It has learned to do so when relevant.
> Owen Amburhttps://www.linkedin.com/in/owenambur/
>  
> 
>     On Thursday, March 13, 2025 at 12:10:21 AM EDT, Paola Di Maio <paola.dimaio@gmail.com> wrote:   
> 
>  Must READ
> 
> Logique & Analyse 183–184 (2003), x–x
> LOGICAL CONSEQUENCE AS TRUTH-PRESERVATION
> STEPHEN READ
> Abstract
> It is often suggested that truth-preservation is insufficient for logical
> consequence, and that consequence needs to satisfy a further condi-
> tion of relevance. Premises and conclusion in a valid consequence
> must be relevant to one another, and truth-preservation is too coarse-
> grained a notion to guarantee that. Thus logical consequence is the
> intersection of truth-preservation and relevance.
> This situation has the absurd consequence that one might concede
> that the conclusion of an argument was true (since the argument had
> true premises and was truth-preserving); yet should refuse to infer
> the conclusion from the premises, in the absence of demonstration
> of the relevance of the premises to the conclusion.
> The error lies in giving insufficient attention to the notion of truth-
> preservation. Relevance is no separable ingredient in the analysis of
> logical consequence, but a necessary condition of it. If an argument
> really is truth-preserving, then that in itself is enough to show that
> the premises are (logically) relevant to the conclus
> https://www.st-andrews.ac.uk/~slr/LogetAnalyse2003.pdf  


Ronald P. Reck

http://www.rrecktek.com - http://www.ronaldreck.com

Received on Thursday, 13 March 2025 15:50:06 UTC