[FRIAM] words

Marcus Daniels marcus at snoutfarm.com
Thu May 30 12:01:34 EDT 2024


I'm not going to watch Joe Rogan, 😊 but I think LLMs don't have to have this homogenous mean problem.  They capture a distribution, so it is a question of the inference procedure to sample from it.  What is the (beam) search algorithm, how deep does it go, and what is the sampling temperature.  

-----Original Message-----
From: Friam <friam-bounces at redfish.com> On Behalf Of glen
Sent: Wednesday, May 29, 2024 1:09 PM
To: friam at redfish.com
Subject: [FRIAM] words

Terrence Howard | Full Address and Q&A | Oxford Union https://youtu.be/ca1vIYmGyYA?si=vhbtA5WUX1CV8LZH

Joe Rogan Experience #2152 - Terrence Howard
https://youtu.be/g197xdRZsW0?si=kFTa7lQJI1lKA6R1

I just can't help but wonder how many people, while listening to Howard talk, realize they're interacting with a sick individual (who deserves compassion but does not deserve gullibility). Or how many people are (like Rogan seems to have been) ... uh ... hypnotized by Howard's well-crafted word salad. In this LLM era, where many people, including some on this list, are enthralled by random bullshit, it seems like a reasonable thing to wonder about. Luckily, the clear cognitive power Howard exhibits puts him in some kind of rare quantile. So our LLMs, being driven mostly to a homogenous mean, their random bullshit will, by definition, match those of us within 1 or a few sigma and suppress the weirdest among us.

Being a fan of steel-manning, I'm having a bit of a crisis. The paradox of tolerance tells me that we absolutely must call bullshit at some point, even if it's not ruthless. Those Oxford Union attendees danced around egging him on and calling him out. Is this what the kids call "cringe"? Do we just cringe and tolerate it? Or, like Rogan, pretend to credibility relying on his weirdness to be so weird that it'll disappear into the tails? Or should we be deplatforming the bullshit?

--
ꙮ Mɥǝu ǝlǝdɥɐuʇs ɟᴉƃɥʇ' ʇɥǝ ƃɹɐss snɟɟǝɹs˙ ꙮ

-. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. .
FRIAM Applied Complexity Group listserv
Fridays 9a-12p Friday St. Johns Cafe   /   Thursdays 9a-12p Zoom https://bit.ly/virtualfriam
to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/
archives:  5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/
  1/2003 thru 6/2021  http://friam.383.s1.nabble.com/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5594 bytes
Desc: not available
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20240530/c2b5c61a/attachment.p7s>


More information about the Friam mailing list