ChatG(eppetto)P(inocchio)T
« previous post | next post »
The start of today's SMBC:
The last two panels:
The mouseover title: "Poor guy's immortal again."
The aftercomic:
It would be more accurate for the thought balloon to read "Sucks being a stochastic stimulus-response machine", but whatever. . .
It's surprising that the Pinocchio==LLM analogy isn't more widespread, especially given the (apparently inevitable) propensity of LLMs for "hallucinations".
Here's what Nano Banana Pro renders given the prompt "Show Sam Altman as Geppetto carving Pinocchio":
Someone with more time than I have this morning should experiment with asking various AI image generation programs to put Sam's head on the original Collodi illustration:
I should also show you what Nano Banana Pro offers as "Show Thinking":
Crafting a Narrative
I'm currently focused on the specific subject. I'm imagining Sam Altman, transformed into Geppetto, deeply engrossed in carving Pinocchio. I need to capture the essence of the scene. I am working on the pose, the setting, and how to best convey the emotion of creation. The style should be realistic, capturing the textures of the wood and the tools.
Developing the Composition
I've broken down the request into key elements. I am visualizing Sam Altman in the role and defining the key action – carving the wooden Pinocchio. Now, I am exploring how to arrange the workshop environment and use lighting to get the desired nostalgic feel. I am thinking of warm, natural light for the desired "candid photograph" vibe.
Verifying the Visuals
I'm now cross-referencing the initial request with the current visual representation. I've focused on the successful integration of the core elements. A specific person is now dressed in specific clothes doing something specific. The focus has been on capturing the desired elements accurately.
Reviewing the Composition
I'm checking that the representation aligns with the user's initial intentions and the detailed description provided. The subject's role, setting, and actions are now precisely rendered and meet the criteria outlined previously. The key focus is on ensuring all the elements match the user's explicit instructions.
Gregory Kusnick said,
December 11, 2025 @ 12:32 pm
It seems possible that Weinersmith wrote "deterministic" rather than "stochastic" because he wasn't actually thinking about LLMs when he wrote it. On its face, the comic is straightforwardly a joke about philosophical zombies, an idea controversial enough on its own to be worth poking fun at.
But it's also possible he was drawing a subtle connection that I would have missed had you not pointed it out.
Kenny Easwaran said,
December 11, 2025 @ 7:38 pm
I've been teaching an AI literacy class several times in the past year, and I assign each of my students to read or watch a classic work of fiction about AI, and have them do various assignments later in the term relating their work of fiction to the topics that are coming up about contemporary AI. Some fictions are more explicitly about robots or computers, like Metropolis or Cat Pictures, Please, and some less so, like The Sorcerer's Apprentice, or Frankenstein, or The Stepford Wives. I've considered having a version of Pinocchio (though not the Steven Spielberg "AI", that had almost been a Kubrick movie) but the issues raised don't quite seem as contemporary as some of the others. (I also try not to include ones where the main point is "AIs go bad and kill people" like Terminator, The Matrix, 2001, Ex Machina, etc.)
Tom said,
December 12, 2025 @ 4:51 pm
@Kenny- Obviously you must be assigning the 1985 film "Weird Science", a true classic of cinema!
KeithB said,
December 16, 2025 @ 7:20 pm
And "The Moon is a Harsh Mistress", with some libertarian philosophy thrown in for free!