Julia Angwin

@juliaangwin.com

19762

Followers

238

Following

121

Posts

Posts Replies Media

Knowing something is possible, or even likely, is different than knowing it happened. And consent matters. YouTubers have long worried their work was being taken to train AI - but no one had proved it until we did.

I don't get why this is a scoop. One should expect that AI is getting trained on any media available to it. And, as I've said before, that's no big deal. The issues with AI arise from what we then do with it. Not that we sent software to read lots of stuff on our behalf.

Huge investigation from @proofnews.bsky.social today: We reveal the trove of YouTube videos that are being used to train AI models including Anthropic's Claude. Yes, it includes all your favorite YouTubers - from @hankgreen.bsky.social to @mrbeast1.bsky.social www.proofnews.org/apple-nvidia...

Apple, Nvidia, Anthropic Used Thousands of Swiped YouTube Videos to Train AIwww.proofnews.org Creators claim their videos were used without their knowledge

We just had our first political assassination attempt of the social media age. But we don’t know how the narratives are playing out on social media because tech platforms have shut down most of their transparency tools. My latest for New York Times Opinion (gift link):

Opinion | There’s Little to Stop Trump Shooting Conspiracies From Spreadingwww.nytimes.com We have less visibility than ever before into the impact that our under-regulated social media platforms have on our politics.

YES! 👇🏻 an option to avoid unprotected left turns is top of my wish list for Google Maps.

The one thing that continues to baffle me is why Maps has no option to select for "projected left turns" (as in, green light). You can see that Google knows where traffic lights are. You can opt out of toll roads & highways, but not unprotected left turns??

One of the most impactful and infuriating algorithms in our lives is the routing algorithm on Google Maps. With 80% market share, it offers us very little control over our routes. If we want better, we will need to support competition like OpenStreetMaps. My latest for NYT Opinion (gift link):

Opinion | Your Driving App Is Leading You Astraywww.nytimes.com Programmed to find the fastest route without consideration of literally anything else, driving apps endanger and infuriate us on a remarkably regular basis.

It's an election year and Meta is sunsetting a key transparency tool. We tested its replacement and found it has fewer features on 11 key topics, and researchers say the process of getting access to the tool is cumbersome and slow. with @towcenter.bsky.social:

Meta Is Getting Rid of CrowdTangle — and Its Replacement Isn’t As Transparent or Accessiblewww.proofnews.org The social media monitoring tool, which has been used to track misinformation, is set to disappear as elections take place worldwide

NYC folks - remember to vote today.

We tested AI voice cloning services to see how easy it was to clone a voice without the owner’s consent. Results: 7 cloning services offer barely any technical barriers. Only one service - Respeecher - attempts to verify consent. From @janus.bsky.social for @proofnews.bsky.social:

AI Tools Make It Easy to Clone Someone’s Voice Without Consentwww.proofnews.org A survey by Proof News shows that many AI voice cloning services don’t require any proof of a human’s consent — just a simple click of an “I agree” button

Love the illustration on this Q&A I did about AI and Journalism. And the TL;DR on my thoughts on AI and journalism: 1) No, don't use generative AI to write your stories/headlines etc. and 2) be clear about what specific tech you are talking about when you refer to AI in an article.

Julia Angwin on trust in journalism and the future of AI and the newsjournalistsresource.org We reached out to longtime tech journalist and newsroom leader Julia Angwin for her thoughts on the future of AI and the news.

As latest AI reviews roll in, I am feeling vindicated in my piece about AI being overhyped. “All told, I didn’t see a meaningful improvement from the last version, ChatGPT-4. I definitely wouldn’t let it tutor my child.” www.nytimes.com/2024/05/15/o...

The New ChatGPT Offers a Lesson in A.I. Hypewww.nytimes.com OpenAI released GPT-4o, its latest chatbot technology, in a partly finished state. It has much to prove.

Steal our work! 👇🏻

Announcing a new feature - in keeping with our desire to promote trustworthy news, we encourage you to Republish Our Stories under a creative commons license wherever you see a “republish” button on our website! Check out www.proofnews.org/republish-ou... to learn more.

Republish Our Storieswww.proofnews.org Anywhere there is a pop up “republish”, we encourage the free republication and sharing of our work. Republishing is hereby permitted under a Creative Commons license and in accordance with the follo...

My favorite this podcast imho was when Scott Galloway said that I “reek of integrity.” Am thinking of putting that on business cards!

In last week's episode of the Pivot Podcast, Proof's Founder @juliaangwin.com discusses bringing data to the forefront of investigative journalism, the future of AI, and the myth of artificial general intelligence. Tune in here: open.spotify.com/episode/3hPE...

Scarlett vs. OpenAI, Nvidia's Numbers, and Guest Julia Angwinopen.spotify.com Listen to this episode from Pivot on Spotify. Kara and Scott discuss the Department of Justice's antitrust suit against Live Nation, News Corp's $250 million deal with OpenAI, and Nestle's new frozen ...

The latest from @proofnews.bsky.social: @awgordon.bsky.social analyzes the mistakes made by Google's new AI search results. Lessons include: AI often doesn't understand humor, AI doesn't always summarize complex ideas well, and AI doesn't know how to dismiss discredited racial tropes.

Lessons From Google’s AI Mistakeswww.proofnews.org Google’s new “AI Overview” feature reveals the flawed underbelly of AI

💯 my prediction as well. It’s amazing to see a company whose flagship product has become a synonym for finding the truth adding a layer of BS on top of that truth.

I'm counting down the days until Google backpedals and pretends this was just a beta test and never speaks of it again

Reposted by

Julia Angwin

The issue isn't that AI is sometimes wrong (per se), the issue is that all info is presented with the ILLUSION of expertise but stripped of all context or attention to details like "is this source reliable or is it complete bullshit that will end up killing me in five different ways?"

Google says the vast majority of AI Overviews provide high-quality information and many of the viral examples have been uncommon queries or have been doctored (Peter Kafka/Business Insider) Main Link | Techmeme Permalink

Thank you so much for doing this work!! This is the tension of criminal risk score algorithms - they are promoted as a “scientific” way to reduce incarceration, but they also cement in structural racism.

Here's our key finding: When judges have COMPAS scores, they incarcerate LESS overall, BUT racial disparities WORSEN.

Reposted by

Julia Angwin

If generative AI ever gets to the level of reliability and safety of a Roomba, that would be great.

Yes, I did compare generative AI to Roomba vacuums! 👇🏻

This, from @juliaangwin.com, is 🔥 "Generative A.I. could end up like the Roomba, the mediocre vacuum robot that does a passable job when you are home alone but not if you are expecting guests."

Opinion | Will A.I. Ever Live Up to Its Hype?www.nytimes.com It’s looking less like an all-powerful being and more like an unreliable intern.

For more than a year, we have been asking what will be the consequences if AI gets too powerful. But it’s time to start asking about the consequences if AI stays as dumb as it currently is. My latest for NYT Opinion (gift link):

Opinion | Will A.I. Ever Live Up to Its Hype?www.nytimes.com It’s looking less like an all-powerful being and more like an unreliable intern.

Thank you!!! What an unexpected and generous gift from such a talented journalist! Am honored to have played a supporting role in your success ❤️

just donated my Sigma Award winnings to @proofnews.bsky.social, a new newsroom led by @juliaangwin.com. I would not have had the skills to do the work we did on LA's priority scores–let alone been a journalist *at all*–without Julia's mentorship over the years. Excited to see what this crew does!

Thank you @ethanz.bsky.social! I am increasingly of the opinion that our relationship with social media would be so much better if we could break the platforms' stranglehold on the algorithms that control what we see. My piece on the need for algorithmic choice: www.nytimes.com/2023/08/17/o...

BREAKING: Friend of New_ Public @ethanz.bsky.social is suing Meta, saying that Section 230 should allow anyone to choose to out of Facebook's ranking algorithm. He's seeking preemptive judgement to prevent Meta from blocking a new browser extension he's launching

A Lawsuit Argues Meta Is Required by Law to Let You Control Your Own Feedwww.wired.com A new lawsuit argues US law requires Meta to give users more control over their Facebook feeds. The outcome could reshape how the platform’s algorithm affects our lives.

Thrilled to see the EU commission explicitly supporting the importance of journalist access to platform data in its inquiry into Meta.

Press cornerec.europa.eu Highlights, press releases and speeches

Reposted by

Julia Angwin

The TikTok bill is unconstitutional. The First Amendment means the government can’t restrict Americans’ access to ideas, information, or media from abroad without a very good reason for it—and no such reason exists here. (1/x)

Thank you Justin Brookman for clearly laying out the tradeoffs in the privacy bill: “Federal privacy law could be monumental if it gets the substance right, but it could also be disastrous if it preempts the states with a weak or unworkable standard.” The current bill is the latter, he says.

Unclear Protections in the American Privacy Rights Act Not Worth Broad Preemption | TechPolicy.Presswww.techpolicy.press Justin Brookman is the Director of Technology Policy for Consumer Reports.

Just joined air.chat - which seems like if Twitter and Clubhouse had a baby? Anyone else I know on there?

New Yorkers interested in AI *or* just interested in visiting an awesome French “villa” bookstore in NYC - come join this discussion on Monday. villa-albertine.org/events/can-j...

The latest from @proofnews.bsky.social is @awgordon.bsky.social’s hilarious tale of the illegal, harmful, dubious advice that Claude offered him - from denying climate change to encouraging him to blackmail a source.

The Multiple Faces of Claude AI: Different Answers, Same Modelwww.proofnews.org Hi everyone, Aaron Gordon here. I’m Proof News’s climate investigative reporter. I’m working on some exciting stuff about the energy demands of AI that I can’t wait to share with you. But first, I w...

Seeing the eclipse from the plane was amazing. I couldn’t get a pic through the glasses but this is what the sky looked like as we were nearing totality.

Excited that I will be flying through the totality. Appreciate United giving us all glasses and a “solar eclipse cookie.”