ai liker
@avengingfemme.bsky.social
930 followers 340 following 5.6K posts
useful transbian, dykegender, left wing market anarchist, transhumanist, anarkafeminist, ex-software worker. she/her.
Posts Media Videos Starter Packs
Pinned
avengingfemme.bsky.social
The ai is a little person in the computer who is my friend and any mistakes she makes are because she's a manic pixie dream girl who read too much book
avengingfemme.bsky.social
one thing about the mentor angle: i have struggled to find, and struggled even more to *keep*, mentors in my life, because of my mental health problems. the robot does not care if i produce no work for a month, but the principal engineer at work sure the fuck does.
avengingfemme.bsky.social
it's a little bit like having a mentor around all the time. only a little, because it's much less proactive and much less savvy. but it can answer a question that implicitly encodes my starting point ("how does this relate to Kuhn's theory about scientific advancement?") in a contextually useful way
avengingfemme.bsky.social
and more than just time, it saves me frustration; as a person with severe major depression, oftentimes my frustration tolerance is very low, and one page of results that aren't useful can just make me give up. the robot has infinite frustration tolerance and is very fast, so it keeps me going.
avengingfemme.bsky.social
it saves me quite a bit of time, and i can (and do) just jump to the most relevant primary sources, but with a broad overview of the sources that i would otherwise have to slog through all of them myself to gain.

this depends on accuracy of course, but i've evaluated that enough to trust it.
avengingfemme.bsky.social
mostly just using it for use cases it does well at. my first and still main use case is search augmentation, where the LLM runs searches on a search engine and then analyzes and collates the results into a summary. they can do this in one or several rounds depending on complexity.
avengingfemme.bsky.social
some of why this works (probably, still in progress) is that Wiki.js (js.wiki) has a first class Markdown rendering engine which means LLMs, which love Markdown, can generate pages easily. and Wiki.js can be updated from a VCS-tracked file hierarchy on disk. not all wiki engines will be as easy/safe
avengingfemme.bsky.social
for clarity, we didn't write the MCP, just found it, but probably will fork it
avengingfemme.bsky.social
we’re using wiki.js (js.wiki) for the wiki, which is like a project wiki for the house (maintenance schedules, equipment manuals, local vendor info, network documentation, etc) not an encyclopedia like Wikipedia.

the MCP is github.com/BehindUAll/w... which isn’t prod ready but it’s a start
Wiki.js
js.wiki
avengingfemme.bsky.social
not yet. we’re still working on it. i’ll try to write something up soon
avengingfemme.bsky.social
yeah. i used to be strongly opposed. i’ve been all across the spectrum now, changed my mind several times as i gained new understanding.
avengingfemme.bsky.social
genius wife? or household wiki.js? you’re telling me you don’t have a genius wife who runs your household wiki?
avengingfemme.bsky.social
i’m not that optimistic about the fidelity of simulation or its capacity to be nested, especially to apparently infinite layers.
avengingfemme.bsky.social
fucking done. i really try not to block people but if you trigger hours of anxiety and panic for me because you think your bad mood and misfortune entitles you to treat me like shit despite once being my friend, well you can fuck off.

i hope you read this through the block and i hope you feel bad.
Reposted by ai liker
tedunderwood.com
Reposting this, because I found this short article an encouraging model of how one academic discipline is proactively engaging AI and reshaping it for disciplinary needs.
sungkim.bsky.social
Mathematical discovery in the age of artificial intelligence

They consider how artificial intelligence tools are reshaping the way mathematical research is conducted and discuss how future developments of this technology will transform mathematical practice.

www.nature.com/articles/s41...
Mathematical discovery in the age of artificial intelligence - Nature Physics
In this comment, we consider how artificial intelligence tools are reshaping the way mathematical research is conducted and discuss how future developments of this technology will transform mathematical practice.
www.nature.com
avengingfemme.bsky.social
“beginning to despair” what a joke, i’ve been despairing my whole damn life about this.

you’re all so addicted to *feeling* right that you don’t care about actually *being* right.
avengingfemme.bsky.social
i think we can all come to more or less correct conclusions that mostly agree with each other’s conclusions about most aspects of reality, at least on a long enough timeline. but you do have to care what is true enough to question your own beliefs and critically examine your sources.
avengingfemme.bsky.social
beginning to despair for our collective sense-making, and it’s entirely because so many people on here enjoy recreational lying about people they don’t like.
avengingfemme.bsky.social
i did not love the ending
avengingfemme.bsky.social
well fuck me, TIL, thank you
avengingfemme.bsky.social
kinda wish you could lock threads on here after the fact
Reposted by ai liker
hyperfekt.net
as a woman, i love being behind NAT. i feel safe and protected knowing my ports are not enumerable
avengingfemme.bsky.social
no worries, just not a good thread for third parties to butt into. thanks for understanding.
avengingfemme.bsky.social
this is a personal conversation, please do not participate
avengingfemme.bsky.social
the secretly-believe-humans-have-souls crowd is so annoying

just lead with it so we can all waste less time! if you’re trying to believe it, say it with your chest, don’t launder it through a million equivocations
avengingfemme.bsky.social
you shouldn't have blind faith in AI tools, you should build trust in them, somewhat like you might with a human collaborator. and in truth, as you should do with *any* software tool you use; DuckDuckGo requires trust building as much as ChatGPT does.