Jenn Wilson
@thatjennwilson.com
PhD Candidate using open LLMs to process complex information for regular people to use. Media, communications, python. Submitting Jan 2027.
Likes lemongrass + ginger tea and chocolate.
Two cats. Australian. They/them.
Likes lemongrass + ginger tea and chocolate.
Two cats. Australian. They/them.
Pinned
A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies | Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems
doi.org
Let's go through a useful paper on anthropomorphic outputs.
DeVrio, Cheng, Egede, Olteanu, and Blodgett published a paper titled "A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies".
It's open access: doi.org/10.1145/3706...
DeVrio, Cheng, Egede, Olteanu, and Blodgett published a paper titled "A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies".
It's open access: doi.org/10.1145/3706...
Reposted by Jenn Wilson
I'm hiring for our interpretable world modeling effort at Percepta! The core question: how can we make an agent build a simulator, understand a company, and then design a policy, all from unstructured data? Come work with me and our amazing team.
jobs.ashbyhq.com/percepta/909...
jobs.ashbyhq.com/percepta/909...
Research Scientist – World Modeling
WHO WE ARE
Percepta’s mission is to transform critical institutions with applied AI. We care that industries that power the world (e.g. healthcare, manufacturing, energy) benefit from frontier techno...
jobs.ashbyhq.com
October 7, 2026 at 3:52 PM
I'm hiring for our interpretable world modeling effort at Percepta! The core question: how can we make an agent build a simulator, understand a company, and then design a policy, all from unstructured data? Come work with me and our amazing team.
jobs.ashbyhq.com/percepta/909...
jobs.ashbyhq.com/percepta/909...
Current nighttime fiction read - there is no anti memetics division by qntm.
Interesting world so far, easy to read prose, good chapter length and pacing.
Interesting world so far, easy to read prose, good chapter length and pacing.
October 7, 2026 at 1:04 PM
Current nighttime fiction read - there is no anti memetics division by qntm.
Interesting world so far, easy to read prose, good chapter length and pacing.
Interesting world so far, easy to read prose, good chapter length and pacing.
Reposted by Jenn Wilson
Thinking more about the point in my undergrad when I became interested to know more rather than do the basics…
This was the year that I was finally old enough to get “youth allowance” from the government. It was the first year I could study without working.
I’m sure that’s not connected at all.
This was the year that I was finally old enough to get “youth allowance” from the government. It was the first year I could study without working.
I’m sure that’s not connected at all.
But thinking on it, I didn't view most of my undergrad like this. It wasn't until my last year when I selected a few really interesting classes (including art history) that the option to choose depth became a genuinely interesting choice.
October 7, 2026 at 5:05 AM
Thinking more about the point in my undergrad when I became interested to know more rather than do the basics…
This was the year that I was finally old enough to get “youth allowance” from the government. It was the first year I could study without working.
I’m sure that’s not connected at all.
This was the year that I was finally old enough to get “youth allowance” from the government. It was the first year I could study without working.
I’m sure that’s not connected at all.
Reposted by Jenn Wilson
im posting this article again terrytao.wordpress.com/2026/09/18/i...
If math is more than proof, we need to better celebrate the rest of it
[This is a guest post by Grant Sanderson. This blog post was initially written in a different file format and converted using AI. — T.] A sentiment echoing throughout the mathematics communit…
terrytao.wordpress.com
October 7, 2026 at 3:10 AM
im posting this article again terrytao.wordpress.com/2026/09/18/i...
I think about this when my friends tell me about students turning in essays at uni that reek of models.
The marker gets paid regardless. The person who loses is the one who is paying for an education they’re not getting.
I have to assume the kids just aren’t aware of what is being traded.
The marker gets paid regardless. The person who loses is the one who is paying for an education they’re not getting.
I have to assume the kids just aren’t aware of what is being traded.
One thing I find endlessly funny about the rise of AI is that it has revealed that there is a large number of people who believe that the reason they wrote essays at school or university is their teachers were desperately waiting for them to provide a definite answer to “Lear: what’s up with him?”
"By using AI, tasks like essays that used to take hours can be completed in minutes!"
Sure, like how if I arrive at the gym and immediately go for a shower, my workout is done in record time. And if I throw my sandwiches straight in the bin, I can finish my lunch in under 10 seconds.
Sure, like how if I arrive at the gym and immediately go for a shower, my workout is done in record time. And if I throw my sandwiches straight in the bin, I can finish my lunch in under 10 seconds.
October 6, 2026 at 10:25 PM
I think about this when my friends tell me about students turning in essays at uni that reek of models.
The marker gets paid regardless. The person who loses is the one who is paying for an education they’re not getting.
I have to assume the kids just aren’t aware of what is being traded.
The marker gets paid regardless. The person who loses is the one who is paying for an education they’re not getting.
I have to assume the kids just aren’t aware of what is being traded.
This is a good point. Often people talk about language models as if only frontier models exist because that's what they hear and talk about.
If you don't want to financially support the frontier labs, you can choose to use open models instead.
Let's talk about the digital divide though.
If you don't want to financially support the frontier labs, you can choose to use open models instead.
Let's talk about the digital divide though.
Something not sufficiently registered in the convo around AI/IP: the value that was gained from pretraining (on our collective legacy) is now common property.
People do pay more for frontier models than open ones — but the differences they're paying for, at this point, come from post-training.
People do pay more for frontier models than open ones — but the differences they're paying for, at this point, come from post-training.
Fortunately, that bridge itself is easy as heck to copy and make free.
At this point pretty much everything that was gained from the initial pre-training is in fact, universally shared. Differences between models that people pay for are now all post training.
At this point pretty much everything that was gained from the initial pre-training is in fact, universally shared. Differences between models that people pay for are now all post training.
October 5, 2026 at 11:10 PM
This is a good point. Often people talk about language models as if only frontier models exist because that's what they hear and talk about.
If you don't want to financially support the frontier labs, you can choose to use open models instead.
Let's talk about the digital divide though.
If you don't want to financially support the frontier labs, you can choose to use open models instead.
Let's talk about the digital divide though.
Standard reminder: you can fork the repository and learn to code. It is an option. C++ is not easy, but it's doable.
Get started today! www.ppsspp.org/docs/develop...
Get started today! www.ppsspp.org/docs/develop...
Dude, this is genuinely so fucking disappointing. Nearly all of the commits I’ve seen for PPSSPP involve generative AI from Claude in the past week alone, and their entire newest dev blog is almost entirely written by Claude. Almost half of the commits in the past 3 months use Claude for AI code.
October 5, 2026 at 12:30 PM
Standard reminder: you can fork the repository and learn to code. It is an option. C++ is not easy, but it's doable.
Get started today! www.ppsspp.org/docs/develop...
Get started today! www.ppsspp.org/docs/develop...
Bevmarks, in Victoria Australia, apparently charges $100 to deliver a pillow locally. I called the store to double check, because I figured something was faulty.
That's the real figure. $100.
I just... what.
That's the real figure. $100.
I just... what.
October 5, 2026 at 5:44 AM
Bevmarks, in Victoria Australia, apparently charges $100 to deliver a pillow locally. I called the store to double check, because I figured something was faulty.
That's the real figure. $100.
I just... what.
That's the real figure. $100.
I just... what.
There are a couple of terms floating around at the moment: cognitive offloading and cognitive surrender. There's also cognitive overload. Yay!
None of these are the same, but they circle around a core idea.
Let's discuss them.
None of these are the same, but they circle around a core idea.
Let's discuss them.
October 1, 2026 at 1:27 AM
There are a couple of terms floating around at the moment: cognitive offloading and cognitive surrender. There's also cognitive overload. Yay!
None of these are the same, but they circle around a core idea.
Let's discuss them.
None of these are the same, but they circle around a core idea.
Let's discuss them.
Once again, Chloe is here to assist with study.
September 29, 2026 at 6:40 AM
Once again, Chloe is here to assist with study.
There’s nothing quite like setting a chapter down for the weekend, allowing your mind to rest, and then coming back to it on Monday where you now can see all of the badly structured statements that your mind concealed from you, because it knew what you meant to say.
September 28, 2026 at 4:06 AM
There’s nothing quite like setting a chapter down for the weekend, allowing your mind to rest, and then coming back to it on Monday where you now can see all of the badly structured statements that your mind concealed from you, because it knew what you meant to say.
Let's go through a useful paper on anthropomorphic outputs.
DeVrio, Cheng, Egede, Olteanu, and Blodgett published a paper titled "A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies".
It's open access: doi.org/10.1145/3706...
DeVrio, Cheng, Egede, Olteanu, and Blodgett published a paper titled "A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies".
It's open access: doi.org/10.1145/3706...
A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies | Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems
doi.org
September 27, 2026 at 3:51 AM
Let's go through a useful paper on anthropomorphic outputs.
DeVrio, Cheng, Egede, Olteanu, and Blodgett published a paper titled "A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies".
It's open access: doi.org/10.1145/3706...
DeVrio, Cheng, Egede, Olteanu, and Blodgett published a paper titled "A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies".
It's open access: doi.org/10.1145/3706...
Reminder: you can be both mad at OpenAI for their decisions and ALSO mad at the Australian Government for their decisions.
www.theguardian.com/australia-ne...
www.theguardian.com/australia-ne...
Albanese says OpenAI hacked Medicare and told Australia months later via email to generic inbox
Prime minister says he told Sam Altman he was disappointed it had taken OpenAI ‘way too long’ to disclose breach
www.theguardian.com
September 24, 2026 at 3:24 AM
Reminder: you can be both mad at OpenAI for their decisions and ALSO mad at the Australian Government for their decisions.
www.theguardian.com/australia-ne...
www.theguardian.com/australia-ne...
The unevenness of access to solar tech irritates me today. Plenty of people on my block are fine right now because they have solar panels. The construction over the back fence is also fine: they have a generator.
Me? Had to throw my half cooked lunch away, and can’t work, because we rent.
Me? Had to throw my half cooked lunch away, and can’t work, because we rent.
September 23, 2026 at 3:21 AM
The unevenness of access to solar tech irritates me today. Plenty of people on my block are fine right now because they have solar panels. The construction over the back fence is also fine: they have a generator.
Me? Had to throw my half cooked lunch away, and can’t work, because we rent.
Me? Had to throw my half cooked lunch away, and can’t work, because we rent.
Blackout in the portion of suburb where I live while I’m working from home. Hoping my draft saved correctly to the cloud before I came downstairs for lunch.
Upside: a few hours to read a book and pat my cat Chloe?
Upside: a few hours to read a book and pat my cat Chloe?
September 23, 2026 at 3:03 AM
Blackout in the portion of suburb where I live while I’m working from home. Hoping my draft saved correctly to the cloud before I came downstairs for lunch.
Upside: a few hours to read a book and pat my cat Chloe?
Upside: a few hours to read a book and pat my cat Chloe?
I'm currently in editing mode for my thesis, in which I ask myself how I can have such good ideas and then write so badly about them.
I think the answer is that I started writing before I was done thinking (because you have to) and so when editing I can *see that process* happen up close. Not good.
I think the answer is that I started writing before I was done thinking (because you have to) and so when editing I can *see that process* happen up close. Not good.
September 22, 2026 at 5:37 AM
I'm currently in editing mode for my thesis, in which I ask myself how I can have such good ideas and then write so badly about them.
I think the answer is that I started writing before I was done thinking (because you have to) and so when editing I can *see that process* happen up close. Not good.
I think the answer is that I started writing before I was done thinking (because you have to) and so when editing I can *see that process* happen up close. Not good.
Discussion further in the thread notes this is about the impact that wfh has on juniors in the workplace - very fair.
WFH removes a lot of opportunity for serendipity in particular, where a small comment or query in a shared kitchen or workspace opens up opportunities that you couldn't predict.
WFH removes a lot of opportunity for serendipity in particular, where a small comment or query in a shared kitchen or workspace opens up opportunities that you couldn't predict.
Really, really hate that the norm is to work from home so much. Trading vibrancy for convenience.
September 19, 2026 at 5:07 AM
Discussion further in the thread notes this is about the impact that wfh has on juniors in the workplace - very fair.
WFH removes a lot of opportunity for serendipity in particular, where a small comment or query in a shared kitchen or workspace opens up opportunities that you couldn't predict.
WFH removes a lot of opportunity for serendipity in particular, where a small comment or query in a shared kitchen or workspace opens up opportunities that you couldn't predict.
Today I am learning about the bears from Katmai National Park so that I may vote in Fat Bear Week.
explore.org/fat-bear-week
explore.org/fat-bear-week
September 19, 2026 at 2:31 AM
Today I am learning about the bears from Katmai National Park so that I may vote in Fat Bear Week.
explore.org/fat-bear-week
explore.org/fat-bear-week
So I mean, fair. But also very funny.
“We want to get to the bottom of who chooses our entrant, on what basis and whether our dismal record of achievement in recent years is the result of bad song choice, weak performance … or something else.”
www.theguardian.com/tv-and-radio...
“We want to get to the bottom of who chooses our entrant, on what basis and whether our dismal record of achievement in recent years is the result of bad song choice, weak performance … or something else.”
www.theguardian.com/tv-and-radio...
UK to answer for its ‘dismal record’ in Eurovision before committee of MPs
Digital, culture, media and sport committee expected to hear evidence from BBC bosses and music industry experts
www.theguardian.com
September 18, 2026 at 11:36 AM
So I mean, fair. But also very funny.
“We want to get to the bottom of who chooses our entrant, on what basis and whether our dismal record of achievement in recent years is the result of bad song choice, weak performance … or something else.”
www.theguardian.com/tv-and-radio...
“We want to get to the bottom of who chooses our entrant, on what basis and whether our dismal record of achievement in recent years is the result of bad song choice, weak performance … or something else.”
www.theguardian.com/tv-and-radio...
Reposted by Jenn Wilson
We invite applications for a new Communications Manager!
This is a supported, 1-year role ideal for an Early Career Researcher, PhD Candidate, or Postdoc who would like to develop their committee, leadership, and communications experience.
Details below. Please share widely!
#dh #comms #ecr
This is a supported, 1-year role ideal for an Early Career Researcher, PhD Candidate, or Postdoc who would like to develop their committee, leadership, and communications experience.
Details below. Please share widely!
#dh #comms #ecr
September 15, 2026 at 2:25 AM
Reposted by Jenn Wilson
be careful if you click this link because the internet archive collection of computer magazines could suck you in for hours archive.org/details/comp...
July 8, 2026 at 4:07 PM
be careful if you click this link because the internet archive collection of computer magazines could suck you in for hours archive.org/details/comp...
Today is “read a pile of literature” day. Chloe is helping.
September 15, 2026 at 12:29 AM
Today is “read a pile of literature” day. Chloe is helping.
Looks like there is an exception for software engineering. I was wondering how they were gonna swing that one.
Note also 15 mins is for kindergarten; it progresses to half lesson for the yr 11 & 12s
education.nsw.gov.au/about-us/edu...
Note also 15 mins is for kindergarten; it progresses to half lesson for the yr 11 & 12s
education.nsw.gov.au/about-us/edu...
In the aftermath to the PISA 2026 scores (where Aus schools actually performed OK for what it is worth), the NSW government have decided to commit to a 'back to books' agenda ... the digital backlash in education is gaining momentum ...
September 13, 2026 at 12:23 PM
Looks like there is an exception for software engineering. I was wondering how they were gonna swing that one.
Note also 15 mins is for kindergarten; it progresses to half lesson for the yr 11 & 12s
education.nsw.gov.au/about-us/edu...
Note also 15 mins is for kindergarten; it progresses to half lesson for the yr 11 & 12s
education.nsw.gov.au/about-us/edu...
Okay, so some upsides:
- they used a temperature of zero (pleasing)
- they used different models (again, good to see)
However:
- Solid agree with Mike, on the dumping everything into the one prompt.
- I'd be wondering about a lost-in-the-middle effect.
- they used a temperature of zero (pleasing)
- they used different models (again, good to see)
However:
- Solid agree with Mike, on the dumping everything into the one prompt.
- I'd be wondering about a lost-in-the-middle effect.
I don't think you should be grading with AI but I also think if you try something like this you've got to do a bit more than dump a scoring rubric into a prompt.
"Researchers have used top Generative #AI models to grade hundreds of undergraduate essays and found that AI only matched human-awarded degree classification around half the time, with AI often failing to assess the best and worst submissions accurately": www.cam.ac.uk/stories/ai-u... #ethics #tech
September 12, 2026 at 7:15 AM
Okay, so some upsides:
- they used a temperature of zero (pleasing)
- they used different models (again, good to see)
However:
- Solid agree with Mike, on the dumping everything into the one prompt.
- I'd be wondering about a lost-in-the-middle effect.
- they used a temperature of zero (pleasing)
- they used different models (again, good to see)
However:
- Solid agree with Mike, on the dumping everything into the one prompt.
- I'd be wondering about a lost-in-the-middle effect.