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
Ted Underwood
@tedunderwood.com
· Nov 4
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
I can't get Marcia Baïla out of my head, since IG has hit me with many videos of it since Catherine Ringer died (french rock 1980s+). This is my fav very weird video of it, though it doesn't have the same angry energy some live perfs had www.youtube.com/watch?v=1zWl...
Les Rita Mitsouko - Marcia Baïla (Clip Officiel)
YouTube video by Les Rita Mitsouko
www.youtube.com
September 20, 2026 at 3:46 PM
I can't get Marcia Baïla out of my head, since IG has hit me with many videos of it since Catherine Ringer died (french rock 1980s+). This is my fav very weird video of it, though it doesn't have the same angry energy some live perfs had www.youtube.com/watch?v=1zWl...
Do people make deals where they go, "This is actually a problem big enough for three papers, so we'll all work equally on all three and each of us will be first-author on one?" Seems like it would be practical.
An advantage of academia over industry is that, untied from competition over profit, we could collaborate freely and widely. In practice, the constraints of first authorship prevent this from being fully realized
September 20, 2026 at 6:23 PM
Do people make deals where they go, "This is actually a problem big enough for three papers, so we'll all work equally on all three and each of us will be first-author on one?" Seems like it would be practical.
very few circumstances where desperate failed dictator cosplay is amusing — and I esp recommend against it when you are in fact a failing dictator without military triumphs to commemorate
September 20, 2026 at 2:35 PM
very few circumstances where desperate failed dictator cosplay is amusing — and I esp recommend against it when you are in fact a failing dictator without military triumphs to commemorate
Reposted by Ted Underwood
wake up it was all a dream, it's 2006 and you are reading a Murakami novel after school longing for something that will never transpire. It's an unusually hot September day, no one is at home; you never heard of embeddings and latent spaces -- and you never will.
September 20, 2026 at 12:06 PM
wake up it was all a dream, it's 2006 and you are reading a Murakami novel after school longing for something that will never transpire. It's an unusually hot September day, no one is at home; you never heard of embeddings and latent spaces -- and you never will.
Reposted by Ted Underwood
People have been creating computer viruses for years, turnabout is fair play
September 20, 2026 at 12:27 PM
People have been creating computer viruses for years, turnabout is fair play
Reposted by Ted Underwood
you have to stop fearmongering about the anthropic biolab btw. this is extremely in-line with what they’ve been saying for like a year and nothing about it is notable or abnormal
September 19, 2026 at 6:20 PM
you have to stop fearmongering about the anthropic biolab btw. this is extremely in-line with what they’ve been saying for like a year and nothing about it is notable or abnormal
Reposted by Ted Underwood
as a longtime textcel I regret to inform that voice input is really good now and you should try it if you haven't recently
September 19, 2026 at 9:46 PM
as a longtime textcel I regret to inform that voice input is really good now and you should try it if you haven't recently
Reposted by Ted Underwood
There's been a lot of this lately, but I thought this one was fairly striking.
Another problem from FrontierMath: Open Problems has been solved! The solution was elicited by Becker, Greger, and Peters in an interactive session with GPT-6 Astra. Peters originally suggested the problem for the benchmark. He had this to say.
September 19, 2026 at 10:45 PM
There's been a lot of this lately, but I thought this one was fairly striking.
Reposted by Ted Underwood
September 19, 2026 at 4:26 AM
Reposted by Ted Underwood
Where I stand on RSI: A moderate's view on the recent events and trajectory of AI. I was underestimating how much we are likely to scale inference-time compute in the near future, but have not seen much to convince me that an intelligence explosion is near.
www.interconnects.ai/p/where-i-st...
www.interconnects.ai/p/where-i-st...
Where I stand on RSI
A moderate's view on the trajectory of AI.
www.interconnects.ai
September 19, 2026 at 3:46 PM
Where I stand on RSI: A moderate's view on the recent events and trajectory of AI. I was underestimating how much we are likely to scale inference-time compute in the near future, but have not seen much to convince me that an intelligence explosion is near.
www.interconnects.ai/p/where-i-st...
www.interconnects.ai/p/where-i-st...
Reposted by Ted Underwood
Good prediction
If we want to update our priors and not just reaffirm them … what do we think Coxon is responding to that he can’t say out loud?
My guess, reading between the lines of “revolutionize any field overnight”: consequential collab between Anthropic and a biotech company that could kick off a land rush.
My guess, reading between the lines of “revolutionize any field overnight”: consequential collab between Anthropic and a biotech company that could kick off a land rush.
September 19, 2026 at 3:34 PM
Good prediction
Reposted by Ted Underwood
huh, that’s new
Here's a chart from the post.... quite a change in a single generation.
September 19, 2026 at 1:57 PM
huh, that’s new
Embarrassed by the baseless speculation I was spreading nine days ago. Should have waited for the base.
September 19, 2026 at 2:20 PM
Embarrassed by the baseless speculation I was spreading nine days ago. Should have waited for the base.
Local newspaper asked me “How important is it to follow tech leaders’ advice to slow AI development? 1 to 10?”
One reason I like talking to local media: it reminds you *what the question is* for most people. I answered “1,” because the advice is self-interested & inequality is a bigger threat.
One reason I like talking to local media: it reminds you *what the question is* for most people. I answered “1,” because the advice is self-interested & inequality is a bigger threat.
September 19, 2026 at 1:55 PM
Local newspaper asked me “How important is it to follow tech leaders’ advice to slow AI development? 1 to 10?”
One reason I like talking to local media: it reminds you *what the question is* for most people. I answered “1,” because the advice is self-interested & inequality is a bigger threat.
One reason I like talking to local media: it reminds you *what the question is* for most people. I answered “1,” because the advice is self-interested & inequality is a bigger threat.
Reposted by Ted Underwood
ask me how I know
there are not enough federal goons to hold a single american city of a million people
I suspect they’re pretty confident in their flock cameras and ICE/CBP goons.
September 19, 2026 at 2:01 AM
ask me how I know
empiricist self-flagellation is my kink, but Opus 5 takes it to a place that makes even *me* uncomfortable
September 18, 2026 at 11:12 PM
empiricist self-flagellation is my kink, but Opus 5 takes it to a place that makes even *me* uncomfortable
Reposted by Ted Underwood
My students! My team! Their smiling faces!
TGIF! Meet the full cast (co-authors) of Data by Design, out October 20, 2026 by @mitpress.bsky.social
September 18, 2026 at 6:05 PM
My students! My team! Their smiling faces!
actually, it’s not so much the thing I possess that matters — it’s the 𝘱𝘳𝘰𝘤𝘦𝘴𝘴
Which all just reeks of what happens when a status order gets upended by some new technique. Basically everyone who is under threat runs around shouting “Never mind what I appeared to value before, what I *really* value is this incredibly intangible thing I possess that also preserves my status”.
September 18, 2026 at 7:00 PM
actually, it’s not so much the thing I possess that matters — it’s the 𝘱𝘳𝘰𝘤𝘦𝘴𝘴
Reposted by Ted Underwood
Transformers are an obscenity you should simultaneously be angry and in awe of how well this brutish fucking thing works.
September 18, 2026 at 4:51 PM
Transformers are an obscenity you should simultaneously be angry and in awe of how well this brutish fucking thing works.
Reposted by Ted Underwood
📢 ARR October 2026 authors
Please review the new ACL Sustainable Reviewing Policy and FAQ to make sure your submissions follow the new requirements.
Policy: www.aclweb.org/portal/conte...
FAQ: aclrollingreview.org/sustainable-...
#ARR #NAACL2027 #COLING2027 #NLProc
Please review the new ACL Sustainable Reviewing Policy and FAQ to make sure your submissions follow the new requirements.
Policy: www.aclweb.org/portal/conte...
FAQ: aclrollingreview.org/sustainable-...
#ARR #NAACL2027 #COLING2027 #NLProc
ACL Sustainable Reviewing Policy | ACL Member Portal
www.aclweb.org
September 18, 2026 at 4:23 PM
📢 ARR October 2026 authors
Please review the new ACL Sustainable Reviewing Policy and FAQ to make sure your submissions follow the new requirements.
Policy: www.aclweb.org/portal/conte...
FAQ: aclrollingreview.org/sustainable-...
#ARR #NAACL2027 #COLING2027 #NLProc
Please review the new ACL Sustainable Reviewing Policy and FAQ to make sure your submissions follow the new requirements.
Policy: www.aclweb.org/portal/conte...
FAQ: aclrollingreview.org/sustainable-...
#ARR #NAACL2027 #COLING2027 #NLProc
Reposted by Ted Underwood
Great to see smart groups chipping away at the integration challenges of open-source AI
(And especially great to see Western governments supporting that work!)
(And especially great to see Western governments supporting that work!)
Big announcement from Montreal yesterday with
@mila-quebec.bsky.social. We're going all in on open source!
blog.mozilla.org/en/mozilla/m...
@mila-quebec.bsky.social. We're going all in on open source!
blog.mozilla.org/en/mozilla/m...
Mila and Mozilla launch new open source AI initiative | The Mozilla Blog
Today at ALL IN, Canada’s largest AI and technology event, Mozilla and Mila announced a new initiative and fresh investment to build an open source AI fo
blog.mozilla.org
September 18, 2026 at 1:35 PM
Great to see smart groups chipping away at the integration challenges of open-source AI
(And especially great to see Western governments supporting that work!)
(And especially great to see Western governments supporting that work!)
Reposted by Ted Underwood
Introducing Benchmark Reviews: our new initiative to audit AI benchmarks. We are launching with 15 benchmarks: 4 Verified, 9 Flawed, and 2 with not enough information for a review.
September 17, 2026 at 9:53 PM
Introducing Benchmark Reviews: our new initiative to audit AI benchmarks. We are launching with 15 benchmarks: 4 Verified, 9 Flawed, and 2 with not enough information for a review.