Proof News

In our latest investigation, Proof News discovered some of the most profitable tech companies on the planet used subtitles from YouTube videos to train their AI models. That includes work from some of their platform's biggest stars, such as Mr Beast and Marques Brownlee. Despite YouTube's policies against unauthorized data scraping, the creators we spoke to had no idea their work had been taken. “It’s theft,” said Dave Wiskus, the CEO of the Nebula, a streaming service partially owned by its creators, many of whom had work taken from YouTube. You can search through the YouTube dataset here: https://lnkd.in/epms6sve https://lnkd.in/ebFg5Bf6

Any journalist trying to find the truth ideally operates a lot like a scientist: start with a hypothesis, examine the evidence, change your mind if the facts change. The New York Times tech contributing editor Julia Angwin founded Proof News to bring that kind of rigor to the coverage of tech. She joins me on the latest In Reality podcast episode to talk AI, democracy, and how to repair social media. Listen! https://lnkd.in/eSuckdcJ

AI has the potential to have a real impact on elections. Julia Angwin grew concerned that no one was putting the leading models to the test to ask basic questions like: → Where should I vote? → Can I vote by text? So Proof News took matters into their own hands. They grilled the top five AI models on essential election info and the results weren’t exactly encouraging. The models fumbled and floundered, spitting out inaccurate answers left and right. But after repeated testing, the models did get better. Not at answering the questions, but at refusing to answer them at all. In the world of AI, sometimes success means knowing when to say "I don't know." We’ll find out soon enough how big the impact of AI will be on our elections... But until then, here’s to hoping more journalists take initiative to do the real digging like Julia and the team at Proof have. — How much faith would you put in AI's election advice? Hear more from Julia, Founder of Proof News, on The Augmented Life, now live at https://lnkd.in/e2Sg3rpF.

“The decision to deprecate the tool will have a direct effect on election integrity experts and their ability to do their work.” In an election year, Meta is sunsetting its transparency tool CrowdTangle. Its replacement? A far less transparent and accessible tool. Read the full joint- investigation by Proof News, Tow Center, and Algorithmic Transparency: https://lnkd.in/eyiruJX5

“The decision to deprecate the tool will have a direct effect on election integrity experts and their ability to do their work.” In an election year, Meta is sunsetting its transparency tool CrowdTangle. Its replacement? A far less transparent and accessible tool. Read the full joint- investigation by Proof News, Tow Center, and Algorithmic Transparency: https://lnkd.in/eyiruJX5

It's an election year and Meta is sunsetting a key transparency tool. Proof News, Tow Center, and The Algorithmic Transparency Institute tested Meta's replacement tool 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. https://lnkd.in/emENwr8A

Janus Rose’s reporting for Proof News found that many AI voice cloning tools lack strict measures to ensure cloning is consensual. The ease with which these tools can replicate voices without permission reflects an industry decision to place the onus for ethical behavior on users, rather than the companies themselves While some companies like Respeecher take a hands-on approach to verifying ethical use of their product many simply require users to check a box asserting they have the right to clone a voice. That’s left voice actors, in particular, worrying about their future as voice cloning has become easy and services ubiquitous. Subscribe to Proof’s newsletter for more reporting that examines the AI industry’s regulation, policy choices and the consequences they may face. https://lnkd.in/gb8GQKqj

Janus Rose’s reporting for Proof News found that many AI voice cloning tools lack strict measures to ensure cloning is consensual. The ease with which these tools can replicate voices without permission reflects an industry decision to place the onus for ethical behavior on users, rather than the companies themselves While some companies like Respeecher take a hands-on approach to verifying ethical use of their product many simply require users to check a box asserting they have the right to clone a voice. That’s left voice actors, in particular, worrying about their future as voice cloning has become easy and services ubiquitous. Subscribe to Proof’s newsletter for more reporting that examines the AI industry’s regulation, policy choices and the consequences they may face. https://lnkd.in/gb8GQKqj

In the latest episode of It’s All Journalism Podcast, Proof News Founder Julia Angwin sat down with Michael O'Connell discuss launching Proof as a news and data-driven journalism studio, the need to restore trust with audiences, and expanding beyond traditional journalism formats. In this conversation Julia details how her background studying math and her love for proofs informs her approach to investigative journalism, her bet on pursuing transparency to build trust with audiences, and how Proof focuses on findings rather than stories. Tune in to The It’s All Journalism Podcast for the full conversation: https://lnkd.in/dZAjq4MK

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 https://lnkd.in/gjYxjnay to learn more.