Lightnews — Scholar-powered news

Ted Underwood

@tedunderwood.com

20K followers 5.6K following 18K posts

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

Posts Media Videos Starter Packs

Pinned

Ted Underwood @tedunderwood.com · Jun 25

New this morning, a Comment I contributed to Nature Computational Science on the interaction between large language models and the humanities. 🧪 🤖 #MLSky

rdcu.be/etk07

The link above will be open-access for a month — plus, I'll reply to this post with a link to a permanently open preprint. +

The impact of language models on the humanities and vice versa

Nature Computational Science - Many humanists are skeptical of language models and concerned about their effects on universities. However, researchers with a background in the humanities are also...

rdcu.be

14 53 160

Reposted by Ted Underwood

Natalie Bennett @nataliegreenpeer.bsky.social · 3h

How little we have understood about the #Biodiversity we are fast trashing: and nature keeps amazing us the more we learn

www.eurasiareview.com/10102025-how...

How Europe’s Largest Bat Catches And Eats Passerines Mid-Air

After nearly 25 years of research, the mystery has finally been solved: Europe’s largest bat doesn’t just eat small birds – it hunts and captures them more than a kilometre above the ground. And it ea...

www.eurasiareview.com

1 11 45

Reposted by Ted Underwood

Sunnie S. Y. Kim ☀️ @sunniesuhyoung.bsky.social · 5h

Our Responsible AI team at Apple is looking for spring/summer 2026 PhD research interns! Please apply at jobs.apple.com/en-us/detail... and email [email protected]. Do not send extra info (e.g., CV), just drop us a line so we can find your application in the central pool!

Machine Learning / AI Internships - Jobs - Careers at Apple

Apply for a Machine Learning / AI Internships job at Apple. Read about the role and find out if it’s right for you.

jobs.apple.com

1 9 16

Ted Underwood @tedunderwood.com · 3h

Welcome to my world

2h
Get a load of this
[blocked]

3h
I just don’t think this is right
[blocked]

4h
If this continues I’m leaving
[blocked]

I never know what ppl are mad about, and I’m ok with that

1 6

Reposted by Ted Underwood

BOO!-rian EEE!-gan @brianckeegan.com · 6h

LaTeX

Merriam-Webster @merriam-webster.com · 1d

What’s the word where you’re from that, when pronounced exactly as it looks, identifies a tourist immediately?

3 9 67

Reposted by Ted Underwood

Nathan Lambert @natolambert.bsky.social · 12h

Talk from Wenting Zhao of Qwen on their plans during COLM. Seems like 1 word is the plan still: scaling training up! Let’s go.

2 2 9

Ted Underwood @tedunderwood.com · 12h

ours, definitely

Ted Underwood @tedunderwood.com · 13h

hmm, I think I don't want to know what you're doing with straws

Reposted by Ted Underwood

Alex Rubinsteyn @alexr.bsky.social · 15h

I’ve been skeptical about AGI = rapid medical progress but starting to believe a weak form of this

(Researchers assisted by the better AIs can move much faster than before, discover far flung areas niches they wouldn’t have considered & get good initial params for experimental design)

3 2 18

Ted Underwood @tedunderwood.com · 14h

wake me when these social media venters have anything that remotely resembles a policy proposal

2 8

Ted Underwood @tedunderwood.com · 14h

No, it's a good idea! We just need something better than DSPy to do the initial variation, I guess.

1 9

Ted Underwood @tedunderwood.com · 14h

You can get 1.5k likes by saying that Excel is terrible, yellow lights should be longer, and bobas should be smaller so they don't get stuck in straws.

excel, bobas, and yellow lights not going anywhere lol

3 13

Ted Underwood @tedunderwood.com · 15h

Yes! And even if they were infinitely patient, you would need a Men in Black memory-wiper to get independent answers repeatedly.

1 13

Reposted by Ted Underwood

Grace @gracekind.net · 15h

Asking Claude for advice 3 times, colorized

1 5 36

Reposted by Ted Underwood

Sebastian Raschka (rasbt) @sebastianraschka.com · 15h

6/
That said, HRM and TRM are fascinating proof‑of‑concepts that show what’s possible with relatively small and efficient architectures. I'm still curious what the real‑world use case will look like. Maybe they could serve as reasoning or planning modules within a larger tool‑calling system.

2 2 9

Ted Underwood @tedunderwood.com · 15h

You can get totally different advice from a language model depending on how you frame your question. Which could be a weakness. On the other hand, if you're patient enough to ask the question several times, with different framing, you start to get a kind of 3D scan of the possibility space.

4 4 71

Reposted by Ted Underwood

SkynetAndChill.com @skynetandchill.com · 19h

Memory and context are a source of lock-in Google and OpenAI won't easily give up.

Carrying Your Personal Memory Across AI Models: A TPM’s Perspective on Persistent Context in a Mult | HackerNoon

Switching between models isn’t the hard part — it’s switching without losing your personal memory that becomes the real challenge.

hackernoon.com

1 2

Reposted by Ted Underwood

Grace Lindsay @neurograce.bsky.social · 17h

What percentage of the benefits of scientific conferences do you think comes from sharing our work with each other vs simply having a deadline we are forced to finish things by? Could be 50/50

4 3 57

Ted Underwood @tedunderwood.com · 18h

Yeah, I am sort of resigned to the idea that there will eventually be a backlash. I can see left-populists and right-populists joining hands on the Protect Children From AI Act or something.

(when they should be joining hands on fighting economic inequality and encouraging open source)

Reposted by Ted Underwood

Maria Antoniak @mariaa.bsky.social · 18h

Keynote at #COLM2025: Nicholas Carlini from Anthropic

"Are language models worth it?"

Explains that the prior decade of his work on adversarial images, while it taught us a lot, isn't very applied; it's unlikely anyone is actually altering images of cats in scary ways.

1 21 73

Ted Underwood @tedunderwood.com · 18h

The Dems were in power 2020-24 though; panic was underway for several years of that. But they (fortunately imo) did little but symbolic regulation

Ted Underwood @tedunderwood.com · 19h

Basically the last sentence here is my point bsky.app/profile/timk...

Tim Kellogg @timkellogg.me · 20h

no not really. one friend is more prone to conspiracy theories but 5G was too far out even for him

with that said, all 3 will still *use* AI, they just complain about the world ending while they do it 😂

Ted Underwood @tedunderwood.com · 19h

I agree but this is consistent with a diagnosis of “skill issue.” Most blog posts a) without an editor b) run 30% too long and c) have a tasteless header image, and a-b-c are related.

Ted Underwood @tedunderwood.com · 19h

Yeah if I had more space I would have said “social media and journalism.”

What I really want to do here is not separate social media from the rest of life, but separate the sphere of opinion from the sphere of action. People grumble about AI irl as well, but they’re not actually trying to stop it.

1 2

Ted Underwood @tedunderwood.com · 1d

I suspect you'll find some of this on Mastodon and some on Facebook too

1 3

Ted Underwood @tedunderwood.com · 1d

yup