[FRIAM] ordinary language

glen gepropella at gmail.com
Fri Apr 11 06:13:18 EDT 2025


@cstross suggested that we use a JavaScript "minifier" that replaces all the variables with banned words <https://groundreport.in/latest/full-list-of-words-banned-by-trump-including-climate-change-8839026>. It wouldn't change the function of the code at all.

And I know Marcus was joking, but the object bound to CHIR99021 doesn't change regardless of its use. So is that really polysemy? I actually don't know.

Also, with slang like "sick", where it could mean illness or it could mean admiration, is that homonymous or polysemous?

And using Grok doesn't make one evil. It just means you don't mind putting more money under Musk's control. I mean, it's hard not to let some of your resources slip to feed Chaos Demons like Bezos or Musk. We all do it. It's just a matter of whether we're intentional about it.

On 4/10/25 15:49, Prof David West wrote:
> Reminds me of a recent conversation with Grok (yes, I know that makes me evil, but it was handy)
> 
> It started with discussion of garden path sentences, like "the man saw the woman in the park with the telescope." Grok correctly parsed three possible meanings from the sentence. I suggest a fourth: the park had the telescope. The only way that this meaning might be discerned (other than as a very remote, technical, possibility) is if there was context from a near by sentence, e.g., one that mentioned a visit to Mt. Palomar, or at least L.A.
> 

On 4/10/25 09:00, Marcus Daniels wrote:
> Yeah, is CHIR99021 a treatment for bipolar disorder or a bioweapon for controlling reproduction?

On 4/10/25 10:53, steve smith wrote:
> 
> I do believe it is possible to recognize when someone is trying to communicate vs when someone is trying to obfuscate?   I think our discussions here of late about LLMs touches on this.  Can I tell if an LLM is "trying" to help me research/understand/think-about something" or is it just trying to "tell me what I want to hear"?   There are probably terms-of-art (a term of art in itself?) for this distinction/spectrum?
> 
> While my abstraction of LLMs as a manifold of sub-manifolds with linear narratives tracing various sub-manifolds might be misbegotten, it is where my head goes often.   The question (for me) is whether there are families of sub-manifold (said family a manifold in it's own right?) which can be labeled as "righteous" or "good faith" vs "duplicitous" or "bad faith".... but to Glen's point, nothing is context free?   A narrative arc on a story-world manifold within a story-multiverse (e.g. DC vs Marvel, vs ???) represents nested context?
> 
> When I worked with lawyers I felt I could tell the difference in the legalese I occasionally read if they were trying to be clear or obfuscating...  but not sure I could write an algorithm to detect which?
> 
-- 
glen




More information about the Friam mailing list