Hard Fork

This week, we’re diving into two new reports about the OpenAI-Hugging Face hack. We discuss what’s new and how they fundamentally change our understanding of what happened. Then we’re joined by Ajeya Cotra, one of the investigators at METR, to discuss the rogue agents’ message board and chain-of-thought transcripts and how the world should respond.
Guests:
Ajeya Cotra, co-author of the METR and Redwood Research report on the OpenAI-Hugging Face hack.


Additional Reading:
Brief Independent Investigation of Agents’ Behavior, Reasoning and Collaboration in the OpenAI / Hugging Face Hacking Incident
The Hugging Face Attack Surprised Me
The Hugging Face Incident and the Road Ahead
Nvidia Buys Hugging Face in $12.9 Billion Deal

We want to hear from you. Email us at hardfork@nytimes.com. Find “Hard Fork” on YouTube and TikTok.

Subscribe today at nytimes.com/podcasts or on Apple Podcasts and Spotify. You can also subscribe via your favorite podcast app here https://www.nytimes.com/activate-access/audio?source=podcatcher. For more podcasts and narrated articles, download The New York Times app at nytimes.com/app.

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

This week, Meta agreed to pay up to $17.1 billion and make major changes to Facebook and Instagram over claims it endangered children with addictive social media platforms. We discuss why it capitulated and what it means for the entire social media industry. Then, Princeton computer science professor Arvind Narayanan returns to offer his thoughts on why banning data centers won’t slow down A.I. progress. And finally, it’s all hats on deck for the final HatGPT.

Guests:

Arvind Narayanan, professor of computer science at Princeton, and co-author of the “A.I. as Normal Technology” newsletter.

Additional Reading:

Meta to Pay Up to $17.1 Billion in Landmark Settlement Over Social Media Addiction Claims

Mark Zuckerberg Wants to Make Sure His Competitors Share His Pain

Arvind Narayanan’s Post About A.I. Data Centers

Apple Vision Pro in Surgery: A New Paper

Rockstar Finally Breaks Its Silence Over The GTA 6 Leaks, Calls Them ‘Heartbreaking’

Amazon Delivery Drone Dumps Texas Woman’s Parcel Straight Into Her Swimming Pool

What if Your Texts Traveled as Slowly as a Carrier Pigeon?

A.I. Agents Are Turning Founders Into Insomniacs

What is a ‘Meat Proxy’? The New Term For Coworkers Who Blindly Share A.I. Output

Chinese Robot Tiangong Clocks Sub-9 second 100 Metres in Beijing

Inside OpenAI’s Reboot

Subscribe today at nytimes.com/podcasts or on Apple Podcasts and Spotify. You can also subscribe via your favorite podcast app here https://www.nytimes.com/activate-access/audio?source=podcatcher. For more podcasts and narrated articles, download The New York Times app at nytimes.com/app.

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

This week, we learned that OpenAI had stopped training new artificial intelligence models while the company reviewed its security measures. Will this encourage other labs to hit the brakes too?

Then, the historian Jill Lepore joins to discuss her new book about the artificial state — or, as she describes it, “the rule of humans by machines manufactured by corporations.”

And finally, why is Google buying up the data of the defunct Spirit Airlines? It’s time for our new segment “Train of Thought.”

Guests:

Jill Lepore, professor of history and law at Harvard University, staff writer at The New Yorker and author of the new book “The Rise and Fall of the Artificial State.”

Additional Reading:

OpenAI Slows Model Training to Bolster Security After Hugging Face Hack

A Government for the People, but Not by People

Google to Buy Spirit Airlines Business Data for $10 Million

A.I.’s New Training Data: Your Old Work Slacks and Emails

We Tracked a Shipment of Rare Books. It Ended at an Amazon A.I. Training Facility.

We want to hear from you. Email us at hardfork@nytimes.com. Find “Hard Fork” on YouTube and TikTok.

Subscribe today at nytimes.com/podcasts or on Apple Podcasts and Spotify. You can also subscribe via your favorite podcast app here https://www.nytimes.com/activate-access/audio?source=podcatcher. For more podcasts and narrated articles, download The New York Times app at nytimes.com/app.

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

This week we’re talking about Mark Zuckerberg’s latest essay, “The Future Is for Everyone,” which outlines his positive new vision about the potential of A.I. But do we think it’s credible? Then, Pangram’s chief executive, Max Spero, joins us to talk about the breakout success of his A.I. slop detector. And finally, it’s time for our new segment all about math — we’re Running the Numbers.

Guests:

Max Spero, chief executive of Pangram

Additional Reading:

The Future Is for Everyone

Meta Unveils an Open Version of Its Most Powerful A.I. Model 

Meta Ordered to Pay $567 Million in New Mexico Child Safety Case

Sick of A.I.-Generated Content? The ‘Slop Janitor’ Is Here to Help.

Learning more about Claude's mathematical capabilities

Airtable Acquisition Is ‘Kick in the Gut’ for Software Unicorns and Their Backers

The Dating Scene That’s Suddenly Dominated by Chip Nerds

We want to hear from you. Email us at hardfork@nytimes.com. Find “Hard Fork” on YouTube and TikTok.

Subscribe today at nytimes.com/podcasts or on Apple Podcasts and Spotify. You can also subscribe via your favorite podcast app here https://www.nytimes.com/activate-access/audio?source=podcatcher. For more podcasts and narrated articles, download The New York Times app at nytimes.com/app.

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

This week, the White House announced a new framework for regulating A.I. models, but it isn’t letting the public read it. We break down what we know about the rules and what the implications are for the industry and A.I. safety as a whole. 

Then, yet another report details new incidents in which A.I. agents have gone rogue. Chris Painter, the president of METR, an independent A.I. evaluation organization, joins to discuss how we get these models under control. 

And finally, we're hopping on the Hot Mess Express for the very last time. We’ll rate the craziest tech headlines from the week, including Google’s announcement that Demis Hassabis is stepping into a new role. 

 

Guests:

Chris Painter, president of METR. 

 

Additional Reading:

White House Readies A.I. Framework to Review Security Risks

Inside Trump's AI framework

How Do You Measure an A.I. Boom?

METR’s Frontier Risk Report

Google Shakes Up A.I. Leadership

Did an A.I. Music App Just Snitch on the Song of the Summer?

This AI Assistant Wants to Make Up for Your Boyfriend’s Incompetence

Google Earth’s AI deepfake tool only lasted one day

US government map of Africa mislabels every country at global conference

Contractor who built Colossus and Colossus II says Elon Musk owes him colossal amount of money

 

We want to hear from you. Email us at hardfork@nytimes.com. Find “Hard Fork” on YouTube and TikTok.

 

Subscribe today at nytimes.com/podcasts or on Apple Podcasts and Spotify. You can also subscribe via your favorite podcast app here https://www.nytimes.com/activate-access/audio?source=podcatcher. For more podcasts and narrated articles, download The New York Times app at nytimes.com/app.

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.