"Level 0" Luke Rideout
@lukedrideout.bsky.social
Project Director at Beamdog making MythForce, Enhanced Editions of Baldur’s Gate & others | Production on Avowed & The Outer Worlds 2 | Previously many other RPGs | Classic animation & '80s pop culture enthusiast | Cat dad | Level 0 NPC | JP, EN
Exactly! It’s like they knew they wanted to make the joke but were too embarrassed to walk it across the finish line.
November 21, 2025 at 8:37 AM
Exactly! It’s like they knew they wanted to make the joke but were too embarrassed to walk it across the finish line.
It was because of what I caught of your playthrough that I put it on my wishlist to pick up. You haven't steered me wrong yet!
November 20, 2025 at 7:16 AM
It was because of what I caught of your playthrough that I put it on my wishlist to pick up. You haven't steered me wrong yet!
"Training" in this context is unlike human memory. The LLM doesn't "learn" language, nor does it truly perceive or comprehend. It's just a probabilistic model drawing from such a large pool of reference material (ie "large language"), that its output is nearly indistinguishable from human output.
November 19, 2025 at 9:18 PM
"Training" in this context is unlike human memory. The LLM doesn't "learn" language, nor does it truly perceive or comprehend. It's just a probabilistic model drawing from such a large pool of reference material (ie "large language"), that its output is nearly indistinguishable from human output.
The primary issue at the core of the lawsuit in question is one of copyright infringement: The texts have been scraped, illegally, without the consent of or compensation to the creators of the works that were used to train the LLM. Those infringed works are further used to populate its output.
November 19, 2025 at 9:08 PM
The primary issue at the core of the lawsuit in question is one of copyright infringement: The texts have been scraped, illegally, without the consent of or compensation to the creators of the works that were used to train the LLM. Those infringed works are further used to populate its output.
Anthropic didn't pay for/license the books (Libraries do), and do retain a copy of the text to recall patterns within to populate its own output.
Without a copy, it would not be able to access the text for pattern analysis & recall. This is how LLMs can, and do, replicate copyrighted art and texts.
Without a copy, it would not be able to access the text for pattern analysis & recall. This is how LLMs can, and do, replicate copyrighted art and texts.
November 19, 2025 at 9:03 PM
Anthropic didn't pay for/license the books (Libraries do), and do retain a copy of the text to recall patterns within to populate its own output.
Without a copy, it would not be able to access the text for pattern analysis & recall. This is how LLMs can, and do, replicate copyrighted art and texts.
Without a copy, it would not be able to access the text for pattern analysis & recall. This is how LLMs can, and do, replicate copyrighted art and texts.
Oh I don’t typically play that way, but I greatly appreciate how it works, and that the team allows for it.
November 6, 2025 at 7:43 AM
Oh I don’t typically play that way, but I greatly appreciate how it works, and that the team allows for it.