No evidence for the current leadership, but the last time we lost multiple major combatants forming the core of the fleet were lost in the opening hours of the war did not end well for the other side.
December 25, 2025 at 9:43 PM
No evidence for the current leadership, but the last time we lost multiple major combatants forming the core of the fleet were lost in the opening hours of the war did not end well for the other side.
Also I'm pretty sure that the Bible doesn't say "God created man in the image of his strength and women in the image of his beauty"
Apparently I missed the bit where the Bible specifies God is a muscle mommy.
Apparently I missed the bit where the Bible specifies God is a muscle mommy.
December 24, 2025 at 9:49 PM
Also I'm pretty sure that the Bible doesn't say "God created man in the image of his strength and women in the image of his beauty"
Apparently I missed the bit where the Bible specifies God is a muscle mommy.
Apparently I missed the bit where the Bible specifies God is a muscle mommy.
Also, you'd need to impose *some* kind of restriction on what the player can ask. Like, if you ask Ulfrik Stormcloak his opinion on US nuclear policy, no you didn't. The story cannot coherently proceed from that question being asked.
December 24, 2025 at 8:07 PM
Also, you'd need to impose *some* kind of restriction on what the player can ask. Like, if you ask Ulfrik Stormcloak his opinion on US nuclear policy, no you didn't. The story cannot coherently proceed from that question being asked.
I could be interested in a Skyrim-like assuming the LLM characters, you know, work. Like, Skyrim's main story is already pretty weak and not what you come for.
But even in that context, I don't trust modern LLMs to stay in character persistently.
But even in that context, I don't trust modern LLMs to stay in character persistently.
December 24, 2025 at 7:32 PM
I could be interested in a Skyrim-like assuming the LLM characters, you know, work. Like, Skyrim's main story is already pretty weak and not what you come for.
But even in that context, I don't trust modern LLMs to stay in character persistently.
But even in that context, I don't trust modern LLMs to stay in character persistently.
It's not even a good Bible paper!
It doesn't cite its Bible references, much less the translation, which is pretty much mandatory for a serious discussion of the Bible
It doesn't cite its Bible references, much less the translation, which is pretty much mandatory for a serious discussion of the Bible
December 24, 2025 at 6:51 PM
It's not even a good Bible paper!
It doesn't cite its Bible references, much less the translation, which is pretty much mandatory for a serious discussion of the Bible
It doesn't cite its Bible references, much less the translation, which is pretty much mandatory for a serious discussion of the Bible
I suspect the fact that a lot of the GDP growth is NVIDIA is decoupling things.
December 23, 2025 at 2:51 PM
I suspect the fact that a lot of the GDP growth is NVIDIA is decoupling things.
Not aware of any President ever naming anything after themselves while in office.
Sometimes stuff gets named after living ex-presidents, but the preference is dead people.
Sometimes stuff gets named after living ex-presidents, but the preference is dead people.
December 22, 2025 at 11:45 PM
Not aware of any President ever naming anything after themselves while in office.
Sometimes stuff gets named after living ex-presidents, but the preference is dead people.
Sometimes stuff gets named after living ex-presidents, but the preference is dead people.
The LLM processing the paper doesn't load it into memory; when an LLM is in inference mode its design is fixed and it doesn't train on input.
It's certainly technically possible to record the inputs and train with them later, but most of the AI companies promise not to do this.
It's certainly technically possible to record the inputs and train with them later, but most of the AI companies promise not to do this.
December 22, 2025 at 11:17 PM
The LLM processing the paper doesn't load it into memory; when an LLM is in inference mode its design is fixed and it doesn't train on input.
It's certainly technically possible to record the inputs and train with them later, but most of the AI companies promise not to do this.
It's certainly technically possible to record the inputs and train with them later, but most of the AI companies promise not to do this.
You can try that, but my understanding is that usually the reporter calls saying "I have a piece going up in ten hours, would you like to comment?" and if you don't get back to them they say "has not responded to a request for comment"
December 22, 2025 at 4:09 AM
You can try that, but my understanding is that usually the reporter calls saying "I have a piece going up in ten hours, would you like to comment?" and if you don't get back to them they say "has not responded to a request for comment"
The games don't go into great detail about the 1950s-2060 period; you see computer records from the eve of the war and there's a DLC set in a holographic recreation of the Chinese invasion of Alaska somewhat earlier.
December 21, 2025 at 11:43 PM
The games don't go into great detail about the 1950s-2060 period; you see computer records from the eve of the war and there's a DLC set in a holographic recreation of the Chinese invasion of Alaska somewhat earlier.
It's basically set in "the future as imagined by the 1950s" rather than our actual future. It's not quite a more technologically advanced 50s; the great communist adversary is China and I'm genuinely not sure Russia existed at the time of the war, and the world is basically out of oil.
December 21, 2025 at 11:38 PM
It's basically set in "the future as imagined by the 1950s" rather than our actual future. It's not quite a more technologically advanced 50s; the great communist adversary is China and I'm genuinely not sure Russia existed at the time of the war, and the world is basically out of oil.
Also, it's bad at distinguishing its instructions from the thing you're asking it to summarize. That mostly doesn't end up being a problem in this use case, but it can have exciting failure modes when it's instructed to search the internet.
December 21, 2025 at 11:28 PM
Also, it's bad at distinguishing its instructions from the thing you're asking it to summarize. That mostly doesn't end up being a problem in this use case, but it can have exciting failure modes when it's instructed to search the internet.
In the worst case, you will discover that it's not actually a summarization machine, it's a text generation machine, and it may decide to generate text that's not from your document. It's been instructed to generate a summary, but it doesn't always listen to instructions.
December 21, 2025 at 11:27 PM
In the worst case, you will discover that it's not actually a summarization machine, it's a text generation machine, and it may decide to generate text that's not from your document. It's been instructed to generate a summary, but it doesn't always listen to instructions.
Its "knowledge" comes from a large corpus, then it uses that to read your document. The rest of the corpus is going to influence how it interprets your document. So even in the best case, the other stuff it's read will influence your summary.
December 21, 2025 at 11:25 PM
Its "knowledge" comes from a large corpus, then it uses that to read your document. The rest of the corpus is going to influence how it interprets your document. So even in the best case, the other stuff it's read will influence your summary.
I did it on some stuff I've been writing as a test, and I wasn't fully satisfied with the result even though I was just looking for an independent summary of about 3k words with no larger context.
December 21, 2025 at 11:22 PM
I did it on some stuff I've been writing as a test, and I wasn't fully satisfied with the result even though I was just looking for an independent summary of about 3k words with no larger context.
I did some AI annotating because I desperately needed the money, and I worked on someone's search-enabled AI.
It basically read the snippets you got on the search result page and wrote from them.
It basically read the snippets you got on the search result page and wrote from them.
December 21, 2025 at 11:18 PM
I did some AI annotating because I desperately needed the money, and I worked on someone's search-enabled AI.
It basically read the snippets you got on the search result page and wrote from them.
It basically read the snippets you got on the search result page and wrote from them.
I stuck around for a while because I had some people I liked following, and I wasn't sure what Mastadon server to join, but most of the people I cared about reading decamped to blusky and I followed.
The people I read who have accounts on both say there's way more real interaction on blusky
The people I read who have accounts on both say there's way more real interaction on blusky
December 21, 2025 at 7:39 PM
I stuck around for a while because I had some people I liked following, and I wasn't sure what Mastadon server to join, but most of the people I cared about reading decamped to blusky and I followed.
The people I read who have accounts on both say there's way more real interaction on blusky
The people I read who have accounts on both say there's way more real interaction on blusky
Though even there your citation should make it clear you haven't read the original; this may come up when the original is in another language you do not read.
December 21, 2025 at 4:50 PM
Though even there your citation should make it clear you haven't read the original; this may come up when the original is in another language you do not read.
When you're citing a paper, it should be because you learned something from the paper. If you didn't read the paper, it has nothing to do with what you're writing.
The only grey area I see is when you read a paper quoting another paper and you want to use the quote but can't read the original.
The only grey area I see is when you read a paper quoting another paper and you want to use the quote but can't read the original.
December 21, 2025 at 4:49 PM
When you're citing a paper, it should be because you learned something from the paper. If you didn't read the paper, it has nothing to do with what you're writing.
The only grey area I see is when you read a paper quoting another paper and you want to use the quote but can't read the original.
The only grey area I see is when you read a paper quoting another paper and you want to use the quote but can't read the original.
I would be inclined to assume the paper is fraudulent and written by AI; a human writer who was fooled by a fake citation in another article would probably cite the article they found it in, not the supposed original.
December 21, 2025 at 4:15 PM
I would be inclined to assume the paper is fraudulent and written by AI; a human writer who was fooled by a fake citation in another article would probably cite the article they found it in, not the supposed original.
Shouldn't, but facial recognition isn't perfect and it's particularly bad with minorities. It's mostly accurate, but the errors are common enough it's not really suitable for making final determinations.
December 21, 2025 at 12:19 PM
Shouldn't, but facial recognition isn't perfect and it's particularly bad with minorities. It's mostly accurate, but the errors are common enough it's not really suitable for making final determinations.
My problem with AI is that it Does Whatever. Like, when I ask it to summarize the nations in The Stormlight Archive, I neither say nor mean "invent a nation of dwarves and giants" but guess what it does.
December 21, 2025 at 1:20 AM
My problem with AI is that it Does Whatever. Like, when I ask it to summarize the nations in The Stormlight Archive, I neither say nor mean "invent a nation of dwarves and giants" but guess what it does.
They seem to have lost sight of the fact that the point of winning elections is to accomplish things. It's not an end unto itself.
December 20, 2025 at 11:20 PM
They seem to have lost sight of the fact that the point of winning elections is to accomplish things. It's not an end unto itself.
Tbh the Lois/Supes interview was the one part I disliked. I just felt Superman should be more secure in his position and have a response to the diplomatic implications he's confident in using. Not necessarily a deeply-thought out one, but one that's convincing to him.
December 20, 2025 at 6:21 PM
Tbh the Lois/Supes interview was the one part I disliked. I just felt Superman should be more secure in his position and have a response to the diplomatic implications he's confident in using. Not necessarily a deeply-thought out one, but one that's convincing to him.