Anna Koop
banner
akoopa.bsky.social
Anna Koop
@akoopa.bsky.social
Eternal learner, incorrigible thinker. Roaming AI researcher, maker, idea generator. Sometimes I even complete things. Eventually. Best known for infectious enthusiasm and irrepressible outbursts. My laughter is with me, not at you. Cis settler ally.
The arrogance towards interesting humanities research is soooo frustrating. STEM isn’t everything.
Tell me you don't understand the purpose of humanities research without telling me you don't understand the purpose of humanities research

@economist.com @eui-history.bsky.social

www.economist.com/europe/2025/...
September 30, 2025 at 2:29 PM
Tech women! My friend is doing a study on women in tech and information systems. Interview and a few diary entries, easy. All tech connections welcome.
🪅🪅🪅 I am recruiting participants for my doctoral research project. Please share widely. ✨✨✨
September 30, 2025 at 2:28 PM
Reposted by Anna Koop
Please contribute if you're able or share the original if you're not.

Note that the most effective way to get help for people is to directly ask people you know
UPDATE NOUR ⚠️⚠️⚠️

She is pregnant
Her husband is sick and recently underwent surgery

„I do not want to die here.
I need your help urgently.
I live in a tower.
Two days ago, 50 residential
towers were bombed.
I need your help to survive.“

Emergency Relief for Ziad & Julia
tinyurl.com/HelpNourFamily
September 23, 2025 at 10:03 PM
Reposted by Anna Koop
we have gone from a world in which we were told not to cite wikipedia because it was unreliable to a world where wikipedia might be the only reliable source left on the internet and we all owe a lot to the pedantic nerds who got us there
for all people mock it wikipedia is genuinely one of the wonders of the modern world
September 21, 2025 at 8:52 PM
Reposted by Anna Koop
If you’re on academia dot edu, let me suggest that you strongly consider deleting your account.
September 17, 2025 at 7:09 PM
Reposted by Anna Koop
The debate around what is or isn’t “cancel culture” really just obscures a simple truth that it’s BAD to be massively racist/sexist/homophobic/transphobic and it’s GOOD to speak out against injustice. Some things are good and some things are bad and it’s okay to say which you think are which.
September 20, 2025 at 4:28 PM
The plausible bullshit matching reality is a *uncontrollable and inconsistent* side effect.
September 21, 2025 at 12:30 PM
We can care for each other.
Kindness saves lives.

Choose kindness every time. 💗
August 31, 2025 at 3:15 PM
It’s got to be terrifying as parents, knowing there’s a treatment that can make a lifelong difference and needing help to make it.
August 31, 2025 at 3:14 PM
I like the monocropping parallel. And it hadn’t occurred to me before how doing eugenics is essentially saying that the genes you prefer can’t win without help.

Like the whole anti-woke movement is straight up admitting you can’t thrive unless you keep everybody else down.
eugenics is genetic monocropping. by no means does it prove "natural" genetic fitness. a program of eugenics does not see which genes outcompete others, rather such a program dominates & subjugates the environment (land and market) in artificial, unsustainable favor of certain gene expressions
August 30, 2025 at 4:19 PM
Reposted by Anna Koop
Data are necessarily biased because they consists of observations of particular phenomena taken by specific people, from specific vantage points, using specific instruments, for a specific purpose. Data are literally the encoding of particular biases—that’s what makes it data.
I've had people tell me to my face that AI helps avoid biases and I haven't laughed harder in a long time.

Data doesn't grow on trees. We decide who matters and what to quantify about them. The very process bakes in bias.
2. AI enhances inclusion. Maybe not. AI systems rely on datasets based on mainstream school populations "and often fail to adequately include children with complex, atypical or multiple needs.

As a result, learners with SEND may be statistically invisible"
schoolsweek.co.uk/ai-bias-pose...
August 30, 2025 at 3:03 PM
It is *tremendously* irresponsible of the article to use the word “persuasion” for interacting with a plausible bullshit generator. It does not “believe” anything. It’s even worse than simply statistical word frequency - it’s that skewed towards what people wanted to hear in artificial tests.
You are not persuading the text pipeline! You are stumbling on to a mathematical vector in which you are causing the word delivery mechanism to replicate sequences other people typed up in which they are persuaded and that OpenAI likely stole without permission!
A study focused on OpenAI's GPT-4o mini found that LLMs can be persuaded to comply with objectionable requests using the same tactics that persuade humans (Dina Bass/Bloomberg)

Main Link | Techmeme Permalink
August 30, 2025 at 2:41 PM
Reposted by Anna Koop
Huh maybe if they weren't kicking all the Palestinians off the app there would be more users

Maybe if they didn't welcome Jesse Singal there would be more users
Bluesky has a growth problem

Bluesky has hit 38M monthly users, up 8M since March

Once adding 5M users per month, it’s now slowed to 1.6M. A 60% drop

But post volume is falling—more users, less activity

It risks becoming a ghost town with a loyal but shrinking core

#bluesky #fediverse
Bluesky Reaches 38 Million Users, Though Post Volume is Declining
Bluesky’s growth momentum has slowed, but it is still growing.
www.socialmediatoday.com
August 7, 2025 at 2:59 PM
Oh this is true in work too. It’s amazing how debilitating it is once you start to suspect your workplace of demanding meaningless work.
My philosophy and experience with students is that they can almost always do incredibly difficult work, but if you ever waste their time with soul-killing meaningless work, you have essentially poisoned the well when it comes to getting them to spend any additional effort on your class.
August 5, 2025 at 12:11 PM
Reposted by Anna Koop
Billionaires could solve world hunger but instead they are making a Ring to enslave Middle-earth.
August 5, 2025 at 12:07 PM
Even here we have people getting urgent care while meeting basic needs can’t be done. I can’t imagine how it feels in Palestine, having only limited glucose solution to treat ongoing starvation.

Everyone needs food. Everyone should have food.
August 3, 2025 at 3:10 PM
Dr Menna’s campaign page now has a note reassuring donors that if the entire family is lost the money will go to help others. 💔

What a thing to have to consider, how *I* will feel about my money after *you and your entire family* is murdered!!
August 3, 2025 at 3:02 PM
Reposted by Anna Koop
Please share as wide as possible, don’t just ‘like’ this message. Lives depend on making sure as many get this as possible. #Tsunami #Pacific #Earthquake #DemVoice1 www.cbc.ca/news/canada/...
Tsunami advisory issued for several parts of the B.C. coast after major earthquake near Russia | CBC News
A tsunami watch has been issued for multiple parts of the B.C. coast after a large earthquake in the Pacific basin near Kamchatka, Russia, on Tuesday afternoon.
www.cbc.ca
July 30, 2025 at 1:57 AM
Reposted by Anna Koop
Problem with writing, or any art, in an information-commodity economy is that you're trying to pursue contradictory goals that in a better world would be aligned: art should be free; artists should have the means to live securely and with a modicum of comfort. Above all with free time to make art.
July 29, 2025 at 11:44 PM
Reposted by Anna Koop
At a time when the world can seem so grim, @danmcquillan.bsky.social’s work on decomputing can provide a ray of hope.

We don’t just need to reclaim control over technology, but of our collective vision of the future.
The tech industry has captured our imaginations, in part because we’ve lost a vision for a better future.

@danmcquillan.bsky.social explains how decomputing isn’t just about exerting control over technology; it’s about imagining a better world.

Full ep: techwontsave.us/episode/286_...
July 28, 2025 at 8:41 PM
Reposted by Anna Koop
The other 50% is iteration. GenAI is TERRIBLE at iteration in general, much less precise iteration. We're astonishingly far from prompting, "use what you've generated as a new basis and just change [element]" and getting anything useful back.

It's just not how LLM dot-connection works natively.
July 29, 2025 at 4:45 AM
Reposted by Anna Koop
A trans person being in the same room as you is not a violation of your privacy. Demanding to know whether a person in a room with you is trans or not IS a violation of their privacy. It's endemic of how fucking far the needle has moved that this simple fact is so lost on so many people.
July 29, 2025 at 9:18 AM
Reposted by Anna Koop
If someone asks a question about "AGI", the only reality-based response is to demand a definition of "AGI" before replying. #acl2025
July 28, 2025 at 8:28 AM