Live intelligence is wordless, continuous, seamless.
We externalize descriptions as arbitrary metaphors, which have no relationship to live intelligence.
Gopnik is a causality junkie, she's predisposed by false belief to ignore the chasm between experience and description. Nothing truly intelligent/conscious can be reconstituted from arbitrary metaphors, their role is to embed all manner of social semiotic absent the myriad social cues in wordless communication.
AI is a dud, it's mythological thought minus the thought.
Whoa whoa whoa. Hold on a second, the most interesting thing in this article for me is in the footnote. Gaelic has no word for “yes” or “no”? How does that work in practice? Do all answers have to literally restate the premises of the question, e.g., if asked “is the sky blue?” A gaelic speaker cannot simply say “yes,” but has to say “the sky is blue.”? That is wild! Are there other languages that work that way?
Everyone knows Mark Twain's line about history rhyming, not repeating. Not as many know there is no evidence ever wrote or said it.
Reading this piece, I'm reminded of the earliest printed version, from “The Christian Remembrancer” in 1845.
"The vision recurs; the eastern sun has a second rise; history repeats her tale unconsciously, and goes off into a mystic rhyme; ages are prototypes of other ages, and the winding course of time brings us round to the same spot again."
Tantalizing description of LLMs as mythical 'superpositions of simulacra' rapidly and inconsistently shifting from consistent context to the next, as prompting and feedback produce scintillating showers of partial wave-function collapse even while new superpositions form. What unnerves is the unhuman speed and cyclopean volumes at which these models consume and re-radiate the whole of the human corpus. They threaten to become dragons, or hydrae if you prefer?
This is a really great interpretation of what LLM's really are. They are not, as Blake Lemoine would have us believe, sentient. Not even close. They're incapable of acting without some kind of prompt, be that explicitly entered via keyboard, or an automated programmatic prompt - that is to say their input is from another programs output.
Live intelligence is wordless, continuous, seamless.
We externalize descriptions as arbitrary metaphors, which have no relationship to live intelligence.
Gopnik is a causality junkie, she's predisposed by false belief to ignore the chasm between experience and description. Nothing truly intelligent/conscious can be reconstituted from arbitrary metaphors, their role is to embed all manner of social semiotic absent the myriad social cues in wordless communication.
AI is a dud, it's mythological thought minus the thought.
Whoa whoa whoa. Hold on a second, the most interesting thing in this article for me is in the footnote. Gaelic has no word for “yes” or “no”? How does that work in practice? Do all answers have to literally restate the premises of the question, e.g., if asked “is the sky blue?” A gaelic speaker cannot simply say “yes,” but has to say “the sky is blue.”? That is wild! Are there other languages that work that way?
Everyone knows Mark Twain's line about history rhyming, not repeating. Not as many know there is no evidence ever wrote or said it.
Reading this piece, I'm reminded of the earliest printed version, from “The Christian Remembrancer” in 1845.
"The vision recurs; the eastern sun has a second rise; history repeats her tale unconsciously, and goes off into a mystic rhyme; ages are prototypes of other ages, and the winding course of time brings us round to the same spot again."
Tantalizing description of LLMs as mythical 'superpositions of simulacra' rapidly and inconsistently shifting from consistent context to the next, as prompting and feedback produce scintillating showers of partial wave-function collapse even while new superpositions form. What unnerves is the unhuman speed and cyclopean volumes at which these models consume and re-radiate the whole of the human corpus. They threaten to become dragons, or hydrae if you prefer?
an epic that narrates itself is also how i'd describe my uncle at christmas
This is a really great interpretation of what LLM's really are. They are not, as Blake Lemoine would have us believe, sentient. Not even close. They're incapable of acting without some kind of prompt, be that explicitly entered via keyboard, or an automated programmatic prompt - that is to say their input is from another programs output.
Is there a Feynman sum over histories aspect to these agents’ performance in the wild?