Ted Underwood
@tedunderwood.com
Uses machine learning to study literary imagination, and vice-versa. Likely to share news about AI & computational social science / Sozialwissenschaft / 社会科学
Information Sciences and English, UIUC. Distant Horizons (Chicago, 2019). tedunderwood.com
Information Sciences and English, UIUC. Distant Horizons (Chicago, 2019). tedunderwood.com
Pinned
Opinion | AI Is the Future. Higher Ed Should Shape It.
If we want to stay at the forefront of knowledge production, we must fit technology to our needs.
www.chronicle.com
Wrote a short piece arguing that higher ed must help steer AI. TLDR: If we outsource this to tech, we outsource our whole business. But rejectionism is basically stalling. If we want to survive, schools themselves must proactively shape AI for education & research. [1/6, unpaywalled at 5/6] +
Reposted by Ted Underwood
no I'm not "posting through it." it's posting through me
November 28, 2025 at 4:04 AM
no I'm not "posting through it." it's posting through me
Listening to “Upside Down” (Stranger Things, natch) and — how does she make “respectfully I say to thee” sound like it belongs in a pop song from 1980?
November 28, 2025 at 1:22 AM
Listening to “Upside Down” (Stranger Things, natch) and — how does she make “respectfully I say to thee” sound like it belongs in a pop song from 1980?
“Relaxed fit” language model
Plus-sized language model
November 27, 2025 at 11:54 PM
“Relaxed fit” language model
Reposted by Ted Underwood
A thing I like about cybernetics: instead of wringing hands about
WHAT IS THOUGHT, can machines think????
it bypasses the whole conversation and just creates a new category, placing both humans and feedback processing devices inside of it
then gets on with the work of what we get for the combo
WHAT IS THOUGHT, can machines think????
it bypasses the whole conversation and just creates a new category, placing both humans and feedback processing devices inside of it
then gets on with the work of what we get for the combo
November 27, 2025 at 2:07 AM
A thing I like about cybernetics: instead of wringing hands about
WHAT IS THOUGHT, can machines think????
it bypasses the whole conversation and just creates a new category, placing both humans and feedback processing devices inside of it
then gets on with the work of what we get for the combo
WHAT IS THOUGHT, can machines think????
it bypasses the whole conversation and just creates a new category, placing both humans and feedback processing devices inside of it
then gets on with the work of what we get for the combo
Dorothy Pilley was a journalist and mountaineer who in 1921 founded the Pinnacle Club, the first rock-climbing club for women. She also wrote a 60-page letter rejecting a marriage proposal from literary critic I.A. Richards. (A later proposal was eventually accepted.)
November 27, 2025 at 8:08 PM
Dorothy Pilley was a journalist and mountaineer who in 1921 founded the Pinnacle Club, the first rock-climbing club for women. She also wrote a 60-page letter rejecting a marriage proposal from literary critic I.A. Richards. (A later proposal was eventually accepted.)
Enjoying my people’s annual IP Parade, which expresses thankfulness for the intellectual property totems that ensure continuity of our social world.
November 27, 2025 at 5:21 PM
Enjoying my people’s annual IP Parade, which expresses thankfulness for the intellectual property totems that ensure continuity of our social world.
Reposted by Ted Underwood
In light of record submission rates and a large volume of AI-generated slop, SocArXiv recently implemented a policy requiring ORCIDs linked in the OSF profile of submitting authors, and narrowing our focus to social science subjects. Today we are taking two more steps:
/1
/1
November 27, 2025 at 2:54 PM
In light of record submission rates and a large volume of AI-generated slop, SocArXiv recently implemented a policy requiring ORCIDs linked in the OSF profile of submitting authors, and narrowing our focus to social science subjects. Today we are taking two more steps:
/1
/1
Reposted by Ted Underwood
this image was generated with google's latest nana banana pro tool. we have no idea how bad it is about to get.
November 27, 2025 at 10:15 AM
this image was generated with google's latest nana banana pro tool. we have no idea how bad it is about to get.
Reposted by Ted Underwood
And a major open science release from Prime Intellect: they don’t stress it enough but SFT part is beyond post-training. This is a fully documented mid-training with tons of insights/gems on MoE training, asynchronous infra RL, deep research. storage.googleapis.com/intellect-3-...
November 27, 2025 at 7:47 AM
And a major open science release from Prime Intellect: they don’t stress it enough but SFT part is beyond post-training. This is a fully documented mid-training with tons of insights/gems on MoE training, asynchronous infra RL, deep research. storage.googleapis.com/intellect-3-...
Reposted by Ted Underwood
Some sorta ersatz haderach?
November 27, 2025 at 4:29 AM
Some sorta ersatz haderach?
Reposted by Ted Underwood
"I say we take off and nuke the site from orbit. It's the only way to be sure."
November 27, 2025 at 2:30 AM
"I say we take off and nuke the site from orbit. It's the only way to be sure."
Reposted by Ted Underwood
What will the creature made all of seadrift do on the dry sand of daylight; what will the mind do, each morning, waking?
November 26, 2025 at 8:41 PM
What will the creature made all of seadrift do on the dry sand of daylight; what will the mind do, each morning, waking?
Reposted by Ted Underwood
Are you a grad student working on post-1945 culture? Could your research benefit from incorporating some data, even minimally? Want feedback from journal editors?
This Post45 Data Collective virtual workshop may be for you!
Applications are due DECEMBER 1: data.post45.org/news/grad-wo...
This Post45 Data Collective virtual workshop may be for you!
Applications are due DECEMBER 1: data.post45.org/news/grad-wo...
The Post45 Data Collective invites graduate students in the humanities or adjacent fields to explore cultural data reflexively and collaboratively in a mini-workshop hosted virtually on Friday, March 13. Details here: data.post45.org/news/grad-wo...
November 26, 2025 at 6:09 PM
Are you a grad student working on post-1945 culture? Could your research benefit from incorporating some data, even minimally? Want feedback from journal editors?
This Post45 Data Collective virtual workshop may be for you!
Applications are due DECEMBER 1: data.post45.org/news/grad-wo...
This Post45 Data Collective virtual workshop may be for you!
Applications are due DECEMBER 1: data.post45.org/news/grad-wo...
Reposted by Ted Underwood
Great thread of discussion here.
Great piece, Dan. I can confirm: Over the years, this has indeed been a longstanding problem. I've helped fund numerous projects over the years trying to address handwriting recognition and it is exciting to see what we can do today.
New issue of my newsletter: "The Writing Is on the Wall for Handwriting Recognition" — One of the hardest problems in digital humanities has finally been solved, and it's a good use of AI newsletter.dancohen.org/archive/the-...
November 26, 2025 at 5:19 PM
Great thread of discussion here.
I think the answer is “acceptance”:
most thoughtful accounts here have moved toward “AI is real, is happening, and if it has downsides we need to address them.”
However there are still several thousand accounts who hate nothing more than hearing that and will try to stop the process by brigading.
most thoughtful accounts here have moved toward “AI is real, is happening, and if it has downsides we need to address them.”
However there are still several thousand accounts who hate nothing more than hearing that and will try to stop the process by brigading.
Is this platform still massively against AI or has it moved more towards acceptance?
November 26, 2025 at 2:48 PM
I think the answer is “acceptance”:
most thoughtful accounts here have moved toward “AI is real, is happening, and if it has downsides we need to address them.”
However there are still several thousand accounts who hate nothing more than hearing that and will try to stop the process by brigading.
most thoughtful accounts here have moved toward “AI is real, is happening, and if it has downsides we need to address them.”
However there are still several thousand accounts who hate nothing more than hearing that and will try to stop the process by brigading.
Reposted by Ted Underwood
in the before times, I imagined that Language was like clothing or armor, supported from below by the musculature of thought
"let me don my words that you might see the shapes within me"
dust scattered over a ghost, & so showing the shape of that spirit
but these imaginings now seem to me wrong!
"let me don my words that you might see the shapes within me"
dust scattered over a ghost, & so showing the shape of that spirit
but these imaginings now seem to me wrong!
November 26, 2025 at 2:16 PM
in the before times, I imagined that Language was like clothing or armor, supported from below by the musculature of thought
"let me don my words that you might see the shapes within me"
dust scattered over a ghost, & so showing the shape of that spirit
but these imaginings now seem to me wrong!
"let me don my words that you might see the shapes within me"
dust scattered over a ghost, & so showing the shape of that spirit
but these imaginings now seem to me wrong!
Reposted by Ted Underwood
Nothing feels as intimate as getting a reviewer who understood exactly what you were trying to hide
November 26, 2025 at 8:16 AM
Nothing feels as intimate as getting a reviewer who understood exactly what you were trying to hide
Reposted by Ted Underwood
Here at UMass Amherst CICS, we’re searching for TT faculty in NLP – see the link from
www.cics.umass.edu/about/employ...
I’m happy to answer questions of course, too!
www.cics.umass.edu/about/employ...
I’m happy to answer questions of course, too!
Faculty Positions
Open tenure-track and teaching faculty positions in computer science and informatics at the Manning College of Information and Computer Sciences
www.cics.umass.edu
November 26, 2025 at 4:21 AM
Here at UMass Amherst CICS, we’re searching for TT faculty in NLP – see the link from
www.cics.umass.edu/about/employ...
I’m happy to answer questions of course, too!
www.cics.umass.edu/about/employ...
I’m happy to answer questions of course, too!
Reposted by Ted Underwood
Dwarkesh golden response: "I like this idea that the real reward hacking is the human researchers who are too focused on the evals"
🤩🤩🤩
🤩🤩🤩
November 26, 2025 at 3:26 AM
Dwarkesh golden response: "I like this idea that the real reward hacking is the human researchers who are too focused on the evals"
🤩🤩🤩
🤩🤩🤩
Reposted by Ted Underwood
googles-nested-learning-paradigm-could-solve-ais-memory-and-continual venturebeat.com/ai/googles-nes… #AI #memory #Google
November 25, 2025 at 11:35 PM
googles-nested-learning-paradigm-could-solve-ais-memory-and-continual venturebeat.com/ai/googles-nes… #AI #memory #Google
Reposted by Ted Underwood
The thing about being a person on campus who is now associated with Gen AI is that you call someone to talk about not-Gen AI and then the actual conversation must be followed with a 30-min conversation about Gen AI.
There were two things I decided I was going to stay away from. One was Gen AI. I am now deeply embroiled in that topic on campus and it’s taking over my life. The other, climate change (not explicitly related to corporate Gen AI), is now the unavoidable pretext for a future book. Good job, Roopsi.
November 25, 2025 at 9:35 PM
The thing about being a person on campus who is now associated with Gen AI is that you call someone to talk about not-Gen AI and then the actual conversation must be followed with a 30-min conversation about Gen AI.
Reposted by Ted Underwood
🤔💭What even is reasoning? It's time to answer the hard questions!
We built the first unified taxonomy of 28 cognitive elements underlying reasoning
Spoiler—LLMs commonly employ sequential reasoning, rarely self-awareness, and often fail to use correct reasoning structures🧠
We built the first unified taxonomy of 28 cognitive elements underlying reasoning
Spoiler—LLMs commonly employ sequential reasoning, rarely self-awareness, and often fail to use correct reasoning structures🧠
November 25, 2025 at 6:26 PM
🤔💭What even is reasoning? It's time to answer the hard questions!
We built the first unified taxonomy of 28 cognitive elements underlying reasoning
Spoiler—LLMs commonly employ sequential reasoning, rarely self-awareness, and often fail to use correct reasoning structures🧠
We built the first unified taxonomy of 28 cognitive elements underlying reasoning
Spoiler—LLMs commonly employ sequential reasoning, rarely self-awareness, and often fail to use correct reasoning structures🧠
Reposted by Ted Underwood
"Emotions are so simple, it would be cool to map them out in a human understandable way" <<pauses an extremely awkward amount of time>>
—Ilya
—Ilya
November 25, 2025 at 7:01 PM
"Emotions are so simple, it would be cool to map them out in a human understandable way" <<pauses an extremely awkward amount of time>>
—Ilya
—Ilya
Reposted by Ted Underwood
⚠️ Update on Deep Research Tulu (DR Tulu), our post-training recipe for deep research agents: we’re releasing an upgraded version of our example agent, DR Tulu-8B (RL), that matches or beats systems like Gemini 3 Pro & Tongyi DeepResearch-30B-A3B on core benchmarks. 🧵
November 25, 2025 at 7:37 PM
⚠️ Update on Deep Research Tulu (DR Tulu), our post-training recipe for deep research agents: we’re releasing an upgraded version of our example agent, DR Tulu-8B (RL), that matches or beats systems like Gemini 3 Pro & Tongyi DeepResearch-30B-A3B on core benchmarks. 🧵
Reposted by Ted Underwood
non-controversial take
social media is about 1000x worse for accelerating personality disorders than the current form of AI
social media is about 1000x worse for accelerating personality disorders than the current form of AI
November 25, 2025 at 6:41 PM
non-controversial take
social media is about 1000x worse for accelerating personality disorders than the current form of AI
social media is about 1000x worse for accelerating personality disorders than the current form of AI