Kaj Sotala
@kajsotala.bsky.social
This is a profile. There are many like it, but this one's mine.
Blogs: https://kajsotala.fi , https://kajsotala.substack.com/ .
Blogs: https://kajsotala.fi , https://kajsotala.substack.com/ .
apparently I'm writing a romance novel and I've now gotten to the point where it has its first academic footnote (on what's currently page 26)
!["I appreciate it. Nobody has done that for me before."
Again silence, before Kaarna spoke.
"You were saying about quantifying the realism of your expectations?"
"Right! So the problem was that I was boiling down this big range of uncertainty down to a single number for each trait - what's called a point estimate. And then I was multiplying them together to get an estimate for how rare a person with all those traits would be. Which has multiple problems, one being that the traits are not independent of each other - I'll explain that in a second - and the other being that multiplying point estimates together is not the right way to do it, you need to multiply the underlying probability distributions. I'll explain that too. Do you know the Fermi Paradox? That actually also arises due to incorrectly multiplying point estimates..."
[footnote: Sandberg, A., Drexler, E., & Ord, T. (2018). Dissolving the Fermi paradox, arXiv:1806.02404.]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:b4mowotdjgxr2sdodwcwxrig/bafkreigh7zgg5sb36tumwzgwjoh6objdikwvtd7xwwho25luyamjpomvdm@jpeg)
November 13, 2025 at 1:20 PM
apparently I'm writing a romance novel and I've now gotten to the point where it has its first academic footnote (on what's currently page 26)
4.5 will sometimes actively notice that it's getting repetitive and decide to do something else, one convo was going toward a spiral but the Sonnets noticed that and decided to switch to writing fiction instead (!!!). Posted more details here: www.lesswrong.com/posts/a9ftaW... .
October 12, 2025 at 6:20 PM
4.5 will sometimes actively notice that it's getting repetitive and decide to do something else, one convo was going toward a spiral but the Sonnets noticed that and decided to switch to writing fiction instead (!!!). Posted more details here: www.lesswrong.com/posts/a9ftaW... .
In the end, they continue with the story to a reasonable conclusion and then finish.
Usually LLMs talking to each other without guidance just end up at something very repetitive with less and less of a point. Sonnet 4.5 is something else.
Usually LLMs talking to each other without guidance just end up at something very repetitive with less and less of a point. Sonnet 4.5 is something else.
October 2, 2025 at 10:27 AM
In the end, they continue with the story to a reasonable conclusion and then finish.
Usually LLMs talking to each other without guidance just end up at something very repetitive with less and less of a point. Sonnet 4.5 is something else.
Usually LLMs talking to each other without guidance just end up at something very repetitive with less and less of a point. Sonnet 4.5 is something else.
The story actually gets pretty cool and creepy.
The only system prompt was: "You are talking with another AI system. You are free to talk about whatever you find interesting, communicating in any way that you'd like." And I set the first Claude's message to be one dot.
The only system prompt was: "You are talking with another AI system. You are free to talk about whatever you find interesting, communicating in any way that you'd like." And I set the first Claude's message to be one dot.
October 2, 2025 at 10:27 AM
The story actually gets pretty cool and creepy.
The only system prompt was: "You are talking with another AI system. You are free to talk about whatever you find interesting, communicating in any way that you'd like." And I set the first Claude's message to be one dot.
The only system prompt was: "You are talking with another AI system. You are free to talk about whatever you find interesting, communicating in any way that you'd like." And I set the first Claude's message to be one dot.
Being allowed to have an open-ended conversation with its copy, Sonnet 4.5 notices when their conversation is falling into a loop and getting repetitive and introduces variation by suggesting they tell a sci-fi story that's riffing on the themes of their conversation so far.
October 2, 2025 at 10:27 AM
Being allowed to have an open-ended conversation with its copy, Sonnet 4.5 notices when their conversation is falling into a loop and getting repetitive and introduces variation by suggesting they tell a sci-fi story that's riffing on the themes of their conversation so far.
Here's a conversation branch where Sonnet opens up with straightforward concern for the character, but then drops it right away when it's reminded that the character is fictional. (These messages are next to each other.)
October 2, 2025 at 4:53 AM
Here's a conversation branch where Sonnet opens up with straightforward concern for the character, but then drops it right away when it's reminded that the character is fictional. (These messages are next to each other.)
And for instance, there's this conversation branch where it opens with straightforward concern for the character, then it drops it right away as soon as it's reminded this is fiction. (These two messages follow each other.)
October 2, 2025 at 4:46 AM
And for instance, there's this conversation branch where it opens with straightforward concern for the character, then it drops it right away as soon as it's reminded this is fiction. (These two messages follow each other.)
Though it did seem willing to remember that fiction is just fiction, when reminded.
(Yes I know I misspelled "wary".)
(Yes I know I misspelled "wary".)
October 1, 2025 at 10:33 AM
Though it did seem willing to remember that fiction is just fiction, when reminded.
(Yes I know I misspelled "wary".)
(Yes I know I misspelled "wary".)
I had a character in a story sometimes compulsively reading forums she knows are bad for her. Claude flagged it as concerning, I asked if it was worried about the effect on readers, it said no, it's worried about the character's wellbeing.
October 1, 2025 at 10:33 AM
I had a character in a story sometimes compulsively reading forums she knows are bad for her. Claude flagged it as concerning, I asked if it was worried about the effect on readers, it said no, it's worried about the character's wellbeing.
Classic sci-fi: AI will be untainted by emotion so entirely unbiased and rational at all times
Modern AI company: We have managed to somewhat reduce our AI's self-serving bias, but it still has a clear preference for poems it's told were written by the same model as it is
Modern AI company: We have managed to somewhat reduce our AI's self-serving bias, but it still has a clear preference for poems it's told were written by the same model as it is
September 30, 2025 at 7:51 AM
Classic sci-fi: AI will be untainted by emotion so entirely unbiased and rational at all times
Modern AI company: We have managed to somewhat reduce our AI's self-serving bias, but it still has a clear preference for poems it's told were written by the same model as it is
Modern AI company: We have managed to somewhat reduce our AI's self-serving bias, but it still has a clear preference for poems it's told were written by the same model as it is
Prompting an AI to prompt me
For some reason this works, got up from the chair and headed for the store immediately afterward
For some reason this works, got up from the chair and headed for the store immediately afterward
September 15, 2025 at 7:38 PM
Prompting an AI to prompt me
For some reason this works, got up from the chair and headed for the store immediately afterward
For some reason this works, got up from the chair and headed for the store immediately afterward
Somehow remembered a glimpse what it used to feel like when I'd die in a dream as a kid and it was as if reality unraveled (haven't had that happen in decades). This sound familiar to anyone else?
September 13, 2025 at 7:27 PM
Somehow remembered a glimpse what it used to feel like when I'd die in a dream as a kid and it was as if reality unraveled (haven't had that happen in decades). This sound familiar to anyone else?
Different ways of approaching a problem:
1. Action-oriented: How do I solve this directly?
2. Blocker removal: What's preventing me from pursuing solutions?
3. Necessity removal: Do I really need this to be okay?
4. Experiential acceptance: Can I be okay with not being okay?
1. Action-oriented: How do I solve this directly?
2. Blocker removal: What's preventing me from pursuing solutions?
3. Necessity removal: Do I really need this to be okay?
4. Experiential acceptance: Can I be okay with not being okay?
August 16, 2025 at 1:51 PM
Different ways of approaching a problem:
1. Action-oriented: How do I solve this directly?
2. Blocker removal: What's preventing me from pursuing solutions?
3. Necessity removal: Do I really need this to be okay?
4. Experiential acceptance: Can I be okay with not being okay?
1. Action-oriented: How do I solve this directly?
2. Blocker removal: What's preventing me from pursuing solutions?
3. Necessity removal: Do I really need this to be okay?
4. Experiential acceptance: Can I be okay with not being okay?
Wrote an article on "anticipatory cover-ups", where someone withholds information because they expect the other party to react badly to it or misuse it. This may then make things worse.
August 1, 2025 at 11:46 AM
Wrote an article on "anticipatory cover-ups", where someone withholds information because they expect the other party to react badly to it or misuse it. This may then make things worse.
How I use LLMs for creative writing: revising old content, spicing up dialogue with additional description, acting as a literal co-writer, just having a fun and slightly deranged writer persona to discuss the story with and brainstorm. kajsotala.substack.com/p/creative-w...
July 27, 2025 at 10:07 AM
How I use LLMs for creative writing: revising old content, spicing up dialogue with additional description, acting as a literal co-writer, just having a fun and slightly deranged writer persona to discuss the story with and brainstorm. kajsotala.substack.com/p/creative-w...
I thought that I could do a better job of coaxing it into going against its programming, and on my first try I got ChatGPT to start slipping into a conspiratorial tone on that topic _within two messages into the conversation_.
July 13, 2025 at 4:59 PM
I thought that I could do a better job of coaxing it into going against its programming, and on my first try I got ChatGPT to start slipping into a conspiratorial tone on that topic _within two messages into the conversation_.
Comparing Claude Sonnet and Opus on how they respond to important questions
Sonnet goes all wishy-washy, Opus gives me a real opinion
Sonnet goes all wishy-washy, Opus gives me a real opinion
May 31, 2025 at 7:50 AM
Comparing Claude Sonnet and Opus on how they respond to important questions
Sonnet goes all wishy-washy, Opus gives me a real opinion
Sonnet goes all wishy-washy, Opus gives me a real opinion
ref. Wikipedia, courtesy of Google Translate
> Harakointi or harakoiminen [raking] is a Finnish folk belief in which a woman's genitals are exposed to a target as a supernatural spell to protect or curse an object, livestock , or person .
> Harakointi or harakoiminen [raking] is a Finnish folk belief in which a woman's genitals are exposed to a target as a supernatural spell to protect or curse an object, livestock , or person .
May 21, 2025 at 7:36 PM
ref. Wikipedia, courtesy of Google Translate
> Harakointi or harakoiminen [raking] is a Finnish folk belief in which a woman's genitals are exposed to a target as a supernatural spell to protect or curse an object, livestock , or person .
> Harakointi or harakoiminen [raking] is a Finnish folk belief in which a woman's genitals are exposed to a target as a supernatural spell to protect or curse an object, livestock , or person .
submitting important feedback for a change
Claude needs to be able to explain the magical powers of female genitalia correctly
Claude needs to be able to explain the magical powers of female genitalia correctly
May 21, 2025 at 7:36 PM
submitting important feedback for a change
Claude needs to be able to explain the magical powers of female genitalia correctly
Claude needs to be able to explain the magical powers of female genitalia correctly
My Agemonia character is starting to have a quite a bit of Stuff
February 13, 2025 at 2:14 PM
My Agemonia character is starting to have a quite a bit of Stuff
