Lightnews — Scholar-powered news

sharongoldman.substack.com/p/trumps-100...

Sharon Goldman @sharongoldman.bsky.social · 17d

Trump’s $100,000 H-1B fee is just the latest hurdle for AI talent in the U.S.

In a January article for Fortune tied to the inauguration, I deeply researched the ‘sense of panic’ felt by immigrant AI talent who worried Trump could make an already broken visa system worse

sharongoldman.substack.com

Sharon Goldman @sharongoldman.bsky.social · 17d

Sharon Goldman @sharongoldman.bsky.social · 19d

Many assume C2PA fights deepfakes. But a new report from the World Privacy Forum says it doesn’t — instead it’s building a vast metadata system that raises thorny questions about privacy & who gets to count as ‘trusted.

Thanks to @katekaye.bsky.social for chatting!

fortune.com/2025/09/18/b...

Sharon Goldman @sharongoldman.bsky.social · 21d

It's a podcast world (I regularly think about being at the crazy AI Action Plan rollout a couple of months ago with the All-In bros where JD Vance sat on a couch with them and later the whole thing was rolled away to make way for Trump and his lectern and American flags )

Kara Swisher @karaswisher.bsky.social · 22d

No notes

Sharon Goldman @sharongoldman.bsky.social · Sep 7

"Person says: "So you mean, they're like dumb animals".
Me: "HOW DARE YOU SIR ANIMALS EXHIBIT RICH MORAL SENSIBILITIES BACK AWAY FROM MY OTTERS AND TURTLES" (ok maybe I didn't quite say it that way)"

Yaaaas

Suresh Venkatasubramanian @geomblog.bsky.social · Sep 7

Me, at a NYC public event on AI with @datasociety.bsky.social, in response to a question: "No, AI systems are not moral agents in their own right. They can communicate responses that we can pretend to ascribe moral status to, but they don't possess any moral grounding" 1/2

Inside Anthropic's 'Red Team'—ensuring Claude is safe, and that Anthropic is heard in the corridors of power

Sharon Goldman @sharongoldman.bsky.social · Sep 4

Most AI red teams break models. Anthropic’s Frontier Red Team also evangelizes the risks of AI itself — a rare hybrid of security + policy. I did a deep dive on this unusual group that sits in Anthropic's policy organization under co-founder Jack Clark.

fortune.com/2025/09/04/a...

Unique among AI labs, Anthropic's "Frontier Red Team" reports to its policy chief and has a mandate to publicize the dangers it finds

The MAGA Influencers Rehabilitating Hitler

Reposted by Sharon Goldman

David Brody @dbrody.bsky.social · Sep 3

“Late last year, David Shor, one of the Democratic Party’s top data scientists, surveyed some 130,000 voters about whether they had a “favorable” or “unfavorable” opinion of Jewish people. Hardly anyone over the age of 70 said their view was unfavorable. More than a quarter of those under 25 did.”

A growing constituency on the right wants America to unlearn the lessons of World War II.

www.theatlantic.com

China’s DeepSeek just dropped a new GPT-5 rival — optimized for Chinese chips, priced to undercut OpenAI

Sharon Goldman @sharongoldman.bsky.social · Aug 26

Yes amen exactly this

Will Oremus @willoremus.com · Aug 26

The funny thing is animals actually can suffer and we slaughter them by the billions but sure let’s have an existential panic over whether a software program screams when we turn it off

The Guardian @theguardian.com · Aug 26

Can AIs suffer? Big tech and users grapple with one of most unsettling questions of our times

2 5

Sharon Goldman @sharongoldman.bsky.social · Aug 21

NEW: China’s DeepSeek quietly released an open-source rival to GPT-5 — optimized for Chinese chips and priced to undercut OpenAI. Here's what that means and why that's a big deal. fortune.com/2025/08/21/c...

DeepSeek's new V3.1 model debuts several design innovations and highlights the intensity of the U.S.-China AI rivalry

Meta wants to speed its race to ‘superintelligence’ — but investors will still want their billions in ad revenue

Sharon Goldman @sharongoldman.bsky.social · Aug 20

Superintelligence, shmooperintelligence - for Meta, it's still always going to be about ad revenue, which is, hello, like 98% of revenue 🤷‍♀️

fortune.com/2025/08/20/m...

Even as Meta reorganizes once more to chase AI ‘superintelligence,’ the company’s core business still lives and dies by advertising revenue

Sam Altman’s AI paradox: Warning of a bubble while raising trillions

Sharon Goldman @sharongoldman.bsky.social · Aug 19

I was not invited to Sam Altman’s cozy dinner with reporters in San Francisco last week (whomp whomp), but maybe that’s for the best. I have trouble suppressing exasperated eye rolls when I hear peak Silicon Valley–ironic statements.

fortune.com/2025/08/19/s...

Sam Altman says AI might be in a bubble—then touts trillions for OpenAI’s buildout. Call it the paradox powering Silicon Valley’s biggest spending spree.

AI safety tip: if you don’t want it giving bioweapon instructions, maybe don’t put them in the training data, say researchers

Reposted by Sharon Goldman

Suresh Venkatasubramanian @geomblog.bsky.social · Aug 15

this is very interesting. I do wonder how much the tamper-proofing can survive extreme attacks, but it's encouraging. 1/2

Sharon Goldman @sharongoldman.bsky.social · Aug 15

Thanks to @stellaathena.bsky.social for chatting with me about Deep Ignorance: the new paper/project from Eleuther AI and the UK AISI. Bottom line: Worried AI could teach people to build bioweapons? Don’t teach it how

fortune.com/2025/08/14/w...

New research shows that scrubbing risky material from AI training data can build safeguards that are harder to bypass — and one author calls out tech giants for keeping such work under wraps.

AI safety tip: if you don’t want it giving bioweapon instructions, maybe don’t put them in the training data, say researchers

2 1 2

Sharon Goldman @sharongoldman.bsky.social · Aug 15

Thanks to @stellaathena.bsky.social for chatting with me about Deep Ignorance: the new paper/project from Eleuther AI and the UK AISI. Bottom line: Worried AI could teach people to build bioweapons? Don’t teach it how

fortune.com/2025/08/14/w...

New research shows that scrubbing risky material from AI training data can build safeguards that are harder to bypass — and one author calls out tech giants for keeping such work under wraps.

Why GPT-5’s most controversial feature – the model router – might also be the future of AI

2 12

Sharon Goldman @sharongoldman.bsky.social · Aug 12

NEW: OpenAI's GPT-5 was supposed to be a leap forward. Instead, it sparked backlash, raising questions about whether the model router is the path to AI's future (hint: many say it still is)

fortune.com/2025/08/12/o...

OpenAI’s latest upgrade was supposed to be a leap forward. Instead, it sparked backlash — and raises big questions about whether stitching together multiple models is the path to AI’s future.

https://www.wsj.com/finance/investing/billions-flow-to-new-hedge-funds-focused-on-ai-related-bets-48d97f41?gaa_at=eafs&gaa_n=ASWzDAiB1inpp_vW376mT8ouhsXkO-B5rnY42DVSSYDoyAlzJK2zUBMXlngJIIC_ns8%3D&gaa_...

Sharon Goldman @sharongoldman.bsky.social · Aug 10

Um wasn't Leopold Aschenbremmer focused on AI safety 😱

23-year-old former OpenAI Super alignment team guy running an AI focused hedge fund 😮

t.co/9k28gT4HKd

t.co

3

Sharon Goldman @sharongoldman.bsky.social · Jul 30

Wow - what paper is this from?

Mark Zuckerberg is pouring billions into superintelligence—so why does his Instagram pitch feel so underwhelming?

Sharon Goldman @sharongoldman.bsky.social · Jul 30

Mark Zuckerberg is pouring billions of dollars into AI ‘superintelligence’—so why does today's Instagram pitch feel so underwhelming?

Kind of like a legless metaverse avatar? 💁‍♀️

fortune.com/2025/07/30/m...

The Meta founder and CEO mostly seems to want to sell more Ray Bans.

Podcast, policy, and pomp in D.C. at the surreal celebration of Trump’s AI Action Plan

Sharon Goldman @sharongoldman.bsky.social · Jul 24

NEW: Silicon Valley’s elite descend on D.C. to celebrate Trump’s AI Action Plan in a surreal fusion of podcast and policy

(oh, and Trump ran so long I missed my train back from DC 😂)

fortune.com/2025/07/23/a...

David Sacks, Trump’s AI czar, shared the stage with his cohosts from the All-in podcast on Wednesday, walking through various elements of the new 28-page AI Action Plan.

Sharon Goldman @sharongoldman.bsky.social · Jul 7

I am going to ICML -- happy to meet up! Signal: sharongoldman.43

This former OpenAI researcher thinks we should be gaming out the AI apocalypse

Sharon Goldman @sharongoldman.bsky.social · Jun 26

NEW: I chatted with ex-OpenAI researcher Steven Adler
this week about role-playing a 'rogue' AI (he says we should all try it) 😵‍💫https://fortune.com/2025/06/26/former-openai-researcher-thinks-its-time-to-war-game-ai-superintelligence-scenarios/

Former OpenAI safety researcher Steven Adler explains why he thinks more people should 'wargame' AI superintelligence scenarios.

6 things I learned after spending a week writing a story about Scale AI's Alexandr Wang

Sharon Goldman @sharongoldman.bsky.social · Jun 23

In the billionaires-are-not-relatable category: very weird emailing with Lucy Guo to check that she did, in fact, receive $650M in cash from the Meta/Scale deal and her responding that she did, in fact, receive $750M 😀

open.substack.com/pub/sharongo...

Often one of the great pleasures of writing is the process of researching and crafting a story.

open.substack.com

3

Sharon Goldman @sharongoldman.bsky.social · Jun 5

I was in DC this week. Here’s the takeaway:
Big AI isn’t just knocking on Washington’s door—it’s moved in.

OpenAI models on gov’t supercomputers. Meta’s Llama for defense. Anthropic building tools for national security. And Eric Schmidt? Playing matchmaker.

My latest:
fortune.com/2025/06/05/b...

From Silicon Valley to the Pentagon: Big AI is going full Beltway

In Washington’s new AI universe, national security and corporate AI ambitions are fusing—reshaping policy, procurement, and the global balance of power.

Former Google CEO Eric Schmidt’s AI Expo serves up visions of war, robotics, and LLMs for throngs of tech execs, defense officials, and fresh recruits

Sharon Goldman @sharongoldman.bsky.social · Jun 4

Eric Schmidt’s AI Expo in DC this week showed how fast the AI-military-policy triangle is locking into place. The vision? A future where geopolitical competition is won with frontier models and flying bots. #SCSPAIExpo25

fortune.com/2025/06/04/e...

From drone swarms to doomsday scenarios, former Google CEO Eric Schmidt’s DC showcase reveals how AI, geopolitics, and defense are fusing into one strategic vision.

Former Google CEO Eric Schmidt’s AI Expo serves up visions of war, robotics, and LLMs for throngs of tech execs, defense officials, and fresh recruits

Sharon Goldman @sharongoldman.bsky.social · Jun 4

Spent some time in DC this week: Former Google CEO
Eric Schmidt's AI Expo serves up visions of war, robotics, and LLMs for throngs of tech execs, defense officials, and fresh recruits

fortune.com/2025/06/04/e...

From drone swarms to doomsday scenarios, former Google CEO Eric Schmidt’s DC showcase reveals how AI, geopolitics, and defense are fusing into one strategic vision.