Dan Davis
@bindlestiff.bsky.social
A linguist and software architect for many years, now I create custom guitars and, occasionally, Maloof-style rocking chairs.
LLMs are NOT AI, are addictive, and are worse than useless.
LLMs are NOT AI, are addictive, and are worse than useless.
Truly a solemn thought
November 11, 2025 at 10:02 PM
Truly a solemn thought
I suggest you stop using anthropomorphized jargon. I’m a veteran software architect and I know exactly how they work. Mistaking conceptualization for actual implementation is not a good look.
November 11, 2025 at 9:50 PM
I suggest you stop using anthropomorphized jargon. I’m a veteran software architect and I know exactly how they work. Mistaking conceptualization for actual implementation is not a good look.
It’s objectively faster to press the same button 3 times ¯\_(ツ)_/¯
November 11, 2025 at 9:42 PM
It’s objectively faster to press the same button 3 times ¯\_(ツ)_/¯
Yeah OK dude. Using a euphemism like “knowledge map” doesn’t make it actual knowledge. Semantics is stripped out on input, and WE add it back when we read the responses.
I’ve discussed this at length with true believers like you. Stop wasting my time.
I’ve discussed this at length with true believers like you. Stop wasting my time.
November 11, 2025 at 9:23 PM
Yeah OK dude. Using a euphemism like “knowledge map” doesn’t make it actual knowledge. Semantics is stripped out on input, and WE add it back when we read the responses.
I’ve discussed this at length with true believers like you. Stop wasting my time.
I’ve discussed this at length with true believers like you. Stop wasting my time.
LOL
What is that “semantic representation?”
There are NO semantics in matrices of tokenized graphemes. Meaning is wholly absent in LLM data, and LLMs have no concept whatsoever of semantics.
Don’t be ridiculous. You’re referring to emergent properties, which are just a product of magical thinking.
What is that “semantic representation?”
There are NO semantics in matrices of tokenized graphemes. Meaning is wholly absent in LLM data, and LLMs have no concept whatsoever of semantics.
Don’t be ridiculous. You’re referring to emergent properties, which are just a product of magical thinking.
November 11, 2025 at 9:19 PM
LOL
What is that “semantic representation?”
There are NO semantics in matrices of tokenized graphemes. Meaning is wholly absent in LLM data, and LLMs have no concept whatsoever of semantics.
Don’t be ridiculous. You’re referring to emergent properties, which are just a product of magical thinking.
What is that “semantic representation?”
There are NO semantics in matrices of tokenized graphemes. Meaning is wholly absent in LLM data, and LLMs have no concept whatsoever of semantics.
Don’t be ridiculous. You’re referring to emergent properties, which are just a product of magical thinking.