Semiotics is 1 of the underlying memeplexes of Literary Theory. I had run into it 20 years ago & not made much of it. But when Semiotic Zombies jumped into my head as a description of LLMs, it seemed right.
LLM chatbots give the appearance of being alive - they can answer questions and seem to carry on conversations. People have even been deluded into thinking that some LLMs are conscious. But, they are not alive. So, I think we can characterize them as being ... undead! Like vampires and ... zombies!
So I don't think my intuition was bad.
In fact, this morning, I remembered that, maybe 16-17 years ago, for my birthday a coworker (thanks Jon Anderson!) gave me some zombie-themed stuff (not sure why - maybe Walking Dead had just started???):
- A ZRT (Zombie Response Team) badge, complete with holder and lanyard.
- The best-selling [bs] talking book: "How To Speak Zombie: A Guide for the Living". Press "B" for "Brains".
The book on speaking zombie was insightful, and I remembered the phrase most used by zombies, which encompasses their raison d'etre:
MORE BRAINS![Note, in many traditions, zombies eat the brains of the living.]
So that made me think: what phrase encompasses the raison d'etre of LLMs?
MORE TRAINING DATA!And what is training data? It is content, both word-based and visual, created by humans. It is the distillation of human thoughts, created in a human brain.
So LLMs grow by eating (the products of) human brains => zombies! QED!
And they know nothing of the real world, they live in the world of words & semiotics => semiotic zombies! FTW!
My 1st post on "Semiotic Zombies" was Saturday.
Sunday, this in-depth post from Gary Marcus:
How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI subtitledIn my post "CommonSense™", I talked about how calling LLMs "AI" was totally wrong - they should be considered just another tool in a toolbox.
Neurosymbolic AI is quietly winning. Here’s what that means – and why it took so long
Well duh. That idea has been around forever, and is called "Neurosymbolic AI" - the merger of neural nets & symbol-based systems. Or, as I think it was called in the 1980s, "AI".
Apparently, mostly in wanting a simple message to attract investment $$$, this approach, to me completely sensible, got sidelined in favor of building ever bigger LLMs, with fingers crossed that some day they would cross some threshold & turn into AGI. Yeah right.
Apparently Gary Marcus, who my daughter Erica pointed me at early this year, has been a proponent of Neurosymbolic AI for decades. For this he has been reviled, mocked, vilified, etc. But in this very informational article, he details how the worm is turning.
Geoffrey Hinton is identified as the arch-vilifier; Yann LeCun has turned apostate & said LLMs will never acheive AGI.
This article reminded me how there are people out there for whom AI has been a career - professionals, I think they're called. I'm a citizen scientist who finally got fed up the "AGI any day now" - and the "Full self-driving cars this year" for 12 years - hype cycles & 1st ranted about it just over a year ago. It has been interesting to follow, & Marcus is a great guide. And it is yet another example of how Capitalism Ruins Everything.
This morning, an interesting article in the Lexington Herald-Leader, revisiting the Nazi Grok scandal. I have touched on that here and a couple of days later I linked to Zeynep Tufekci's NYT article:
Artificial Unintelligence,Ha ha, in keeping with the theme of "bullshit", I liked this bit of scatology:
by Jonah Goldberg of The Dispatch.
As far as I understand it, part of the problem is that Grok was initially "trained" like a blue whale. What I mean is that blue whales feed on gazillions of tiny creatures in the ocean. Grok is trained to swim in social media, hoovering up the lingo. And social media is a giant open ocean of garbage. Jew hatred is to social media what plankton is to the ocean.Feed an AI whale a lot of antisemitic swill, and it will defecate antisemitic crap on command.
Goldberg also thoughtfully assesses the RW fetish on "political incorrectness" and what it means.
Then, processing my RSS feeds, another excellent, lengthy article at 404 Media:
The Media's Pivot to AI Is Not Real and Not Going to Work, by Jason Koebler.
So AI is destroying traffic, ripping off our work, creating slop that destroys discoverability and further undermines trust, and allows random people to create news-shaped objects that social media and search algorithms either can’t or don’t care to distinguish from real news. And yet media executives have decided that the only way to compete with this is to make their workers use AI to make content in a slightly more efficient way than they were already doing journalism.Koebler's prescription is to emphasize human products, made for human consumption. That is what I look for. But, like most modern people who aren't retired, I don't think he is going to take the "No LLMs were involved in the creation of this content" pledge. ;->
I like this pic:
Note, this is identified as a photo, which I will take to mean that this pic is not a pledge violation.
Note also, when we have real "AI" people, not deranged zen poets spouting bullshit, I am really looking forward to meeting them. That is part of why the current LLM hype is so disappointing. I will not be a human nativist, again, I look forward to new kinds of people, just as I enjoy all different kinds of people now.
Here's the home page/directory for my posts on Bullshit.
No comments:
Post a Comment