Lightnews — Scholar-powered news

Tejas Srinivasan @tejassrinivasan.bsky.social · Jun 27

🚨Reminder: Submissioms for the ORIGen workshop at COLM are due today!!! 🚨

CfP: origen-workshop.github.io/submissions/

OpenReview submission page: openreview.net/group?id=col...

Tejas Srinivasan @tejassrinivasan.bsky.social · May 16

LLMs are all around us, but how can we foster reliable and accountable interactions with them??

To discuss these problems, we will host the first ORIGen workshop at @colmweb.org! Submissions welcome from NLP, HCI, CogSci, and anything human-centered, due June 20 :)

origen-workshop.github.io

ORIGen 2025

Workshop on Optimal Reliance and Accountability in Interactions with Generative LMs

origen-workshop.github.io

Tejas Srinivasan @tejassrinivasan.bsky.social · May 30

I'm trying to make "bleet" a thing

2 3

Reposted by Tejas Srinivasan

Jesse Thomason @thomason.bsky.social · May 28

This month, @jessezhang.bsky.social completed his PhD defense and signed to start a postdoc with @abhishekunique7.bsky.social at UW! Keep an eye on his journey :) www.jessezhang.net
I'm sad to lose one of my sinistral students but glad to produce another Dr. Jesse 😛

Jesse Thomason and Jesse Zhang in their respective PhD robes.

2 9

Tejas Srinivasan @tejassrinivasan.bsky.social · May 16

The only silver lining of my ACL rejection is that I have something to submit to EMNLP

5

Tejas Srinivasan @tejassrinivasan.bsky.social · May 16

LLMs are all around us, but how can we foster reliable and accountable interactions with them??

To discuss these problems, we will host the first ORIGen workshop at @colmweb.org! Submissions welcome from NLP, HCI, CogSci, and anything human-centered, due June 20 :)

origen-workshop.github.io

ORIGen 2025

Workshop on Optimal Reliance and Accountability in Interactions with Generative LMs

origen-workshop.github.io

4 10

Tejas Srinivasan @tejassrinivasan.bsky.social · May 1

This! So much this!!!

Mark Riedl @markriedl.bsky.social · Apr 29

AI can do so much more. Instead of seeing aging as a problem to sweep under the rug, we should be designing AI to facilitate meaningful connections for all.

1

Reposted by Tejas Srinivasan

Mark Riedl @markriedl.bsky.social · Apr 29

Nothing says “I love you” like outsourcing your parents’ phone calls to a chatbot. 🙃 Social isolation in aging is real. Connection isn’t something you can automate.

Why does everyone think we can just throw a chatbot at every problem?

www.404media.co/i-tested-the...

I Tested The AI That Calls Your Elderly Parents If You Can't Be Bothered

inTouch says on its website "Busy life? You can’t call your parent every day—but we can." My own mum said she would feel terrible if her child used it.

www.404media.co

1 7 42

Tejas Srinivasan @tejassrinivasan.bsky.social · Mar 13

Ty for the plug 🙏
Model confidence is a good decision aid (arxiv.org/pdf/2001.02114), while explanations are less useful and can cause over-reliance (arxiv.org/abs/2310.12558, arxiv.org/pdf/2406.19170). Other interaction cues like AI warmth can also make a difference (arxiv.org/abs/2407.07950).

2

Reposted by Tejas Srinivasan

Jameel Jaffer @jameeljaffer.bsky.social · Mar 9

Arresting and threatening to deport students because of their participation in political protest is the kind of action one ordinarily associates with the world’s most repressive regimes. It’s genuinely shocking that this appears to be what’s going on right here. 1/

36 750 3.1K

Tejas Srinivasan @tejassrinivasan.bsky.social · Mar 10

What do you mean by core capabilities, for VLMS? IMO core capabilities should be determined by the applications we care about, and I'd argue medical use cases are as important (if not more) as MSCOCO-style images/scenes

1 1

Reposted by Tejas Srinivasan

Margaret Mitchell @mmitchell.bsky.social · Mar 6

I worry that concerns with "superintelligence" are being blurred with concerns around *ceding human control*.
A "SuperDumb" system can create mutually assured destruction. What it takes is allowing AI systems to execute code autonomously in military operations.

Margaret Mitchell @mmitchell.bsky.social · Mar 6

AI real talk. We (humanity) are moving full speed ahead at building AI agents for war that can create a runaway missile crisis of mutually assured destruction globally.

Is the option of not allowing AI agents to deploy missiles already off the table, or is that still up for discussion?

Scale AI announces multimillion-dollar defense deal, a major step in U.S. military automation

Spearheaded by the Defense Innovation Unit, the Thunderforge program will work with Anduril, Microsoft and others to develop and deploy AI agents.

www.cnbc.com

3 12 35

Reposted by Tejas Srinivasan

Max Kennerly @maxkennerly.bsky.social · Mar 6

"The first guest on Gavin Newsom's podcast was Charlie Kirk" is more than enough for me to say "absolutely not" to any suggestion Newsom play any role in the future of the Democratic Party. People like him are the past, the failures, the ones who got us here.

Kelsey Atherton @atherton.bsky.social · Mar 5

Gavin Newsom would have put down the Bell Riots with tanks and napalm I can tell you that much

Gavin Newsom v @GavinNewsom
Follow
Make sure to tune in TOMORROW for the first episode of my new podcast
→
linktr.ee/govgavinnewsom...
Charlie Kirk
We had quite the chat!

520 2.2K 11K

Tejas Srinivasan @tejassrinivasan.bsky.social · Feb 28

What are you using o1pro for? And in what aspects do you think it's better than other LLMs?

1

Tejas Srinivasan @tejassrinivasan.bsky.social · Feb 28

Is this advice you reserve for a particular class of problems, or is it just generally applicable because we still don't know the full breadth of LLM capabilities?

1

Tejas Srinivasan @tejassrinivasan.bsky.social · Feb 28

I'm always three days away from being three days away

1

Tejas Srinivasan @tejassrinivasan.bsky.social · Feb 27

We hope our work inspires the community to more closely consider how user characteristics, including but not limited to trust, affect how people rely on AI assistance.

Work done with the always-awesome @thomason.bsky.social!

Tejas Srinivasan @tejassrinivasan.bsky.social · Feb 27

Improving AI reliability is more important than ever as AI systems are increasingly deployed in real-world settings with high stakes. We believe it is important for AI researchers to think about the user-AI dyad 🧑🤖, rather than just the AI in a vacuum.

1

Tejas Srinivasan @tejassrinivasan.bsky.social · Feb 27

These findings show that being able to estimate users’ trust levels can enhance human-AI collaboration 💪 but we also find that modeling user trust is very challenging! 😓 Our work reveals promising new directions for user modeling that extend beyond merely learning user preferences.

1

Tejas Srinivasan @tejassrinivasan.bsky.social · Feb 27

We show that adapting AI behavior to user trust levels, by showing AI explanations during moments of low trust and counter-explanations during high trust, effectively mitigates inappropriate reliance and improves decision accuracy! These improvements are also seen with other intervention strategies.

1 1

Tejas Srinivasan @tejassrinivasan.bsky.social · Feb 27

In two decision-making tasks, we find that low and high user trust levels worsen under-reliance and over-reliance on AI recommendations, respectively 💀💀💀

Can the AI assistant do something differently when user trust is low/high to prevent such inappropriate reliance? Yes!

1

Tejas Srinivasan @tejassrinivasan.bsky.social · Feb 27

People are increasingly relying on AI assistance, but *how* they use AI advice is influenced by their trust in the AI, which the AI is typically blind to. What if they weren’t?

We show that adapting AI assistants' behavior to user trust mitigates under- and over-reliance!

arxiv.org/abs/2502.13321

1 2 8

Reposted by Tejas Srinivasan

Meredith Whittaker @meredithmeredith.bsky.social · Feb 19

Stand by this: www.politico.com/newsletters/...

What’s a technology that you think is overhyped?

I’m going to give a sideways answer to this, which is that the venture capital business model needs to be understood as requiring hype. You can go back to the Netscape IPO, and that was the proof point that made venture capital the financial lifeblood of the tech industry.

Venture capital looks at valuations and growth, not necessarily at profit or revenue. So you don’t actually have to invest in technology that works, or that even makes a profit, you simply have to have a narrative that is compelling enough to float those valuations. So you see this repetitive and exhausting hype cycle as a feature in this industry. A couple of years ago, you would have been asking me about the metaverse, then last year, you would have asked me about Web3 and crypto, and for each of these inflection points there’s an Andreessen Horowitz manifesto.

It’s not simply that one piece of technology is overhyped, it’s that hype is a necessary ingredient of the current business ecosystem of the tech industry. We should examine how often the financial incentive for hype is rewarded without any real social returns, without any meaningful progress in technology, without these tools and services and worlds ever actually manifesting. That’s key to understanding the growing chasm between the narrative of techno-optimists and the reality of our tech-encumbered world.

160 3.2K 9.8K

Tejas Srinivasan @tejassrinivasan.bsky.social · Feb 19

Do each of these correspond to a particular conf deadline? I'm guessing
May: EMNLP
July: AACL?
Oct: EACL/NAACL
Feb: ACL

1

Reposted by Tejas Srinivasan

Mert İnan @merterm.bsky.social · Feb 8

‼️ Ever wish LLMs would just... slow down for a second?

In our latest work, "Better Slow than Sorry: Introducing Positive Friction for Reliable Dialogue Systems", we delve into how strategic delays can enhance dialogue systems.

Paper Website: merterm.github.io/positive-fri...

1 5 14

Reposted by Tejas Srinivasan

Renee DiResta @noupside.bsky.social · Jan 29

“Toward the end of the November dinner, Trump raised the matter of the lawsuit, the people said. The president signaled that the litigation had to be resolved before Zuckerberg could be “brought into the tent,” one of the people said.”

They’re in the tent now. Cowards.

Jeff Horwitz @jeffhorwitz.bsky.social · Jan 29

Meta has agreed to pay Trump $25 million in damages to settle a lawsuit alleging that removing Trump from the platform was illegal. Message was that the money needed to be paid before Meta could be "in the tent." Never got around to filing an amended complaint. www.wsj.com/us-news/law/...

Exclusive | Trump Signs Agreement Calling for Meta to Pay $25 Million to Settle Suit

The president had sued the social-media company after his accounts were suspended.

www.wsj.com

55 240 770