Vue lecture

Il y a de nouveaux articles disponibles, cliquez pour rafraîchir la page.

How Misinformation Spreads? It's Funded By 'The Hellhole of Programmatic Advertising'

Journalist Steven Brill has written a new book called The Death of Truth. Its subtitle? "How Social Media and the Internet Gave Snake Oil Salesmen and Demagogues the Weapons They Needed to Destroy Trust and Polarize the World-And What We Can Do." An excerpt published by Wired points out that last year around the world, $300 billion was spent on "programmatic advertising", and $130 billion was spent in the United States alone in 2022. The problem? For over a decade there's been "brand safety" technology, the article points out — but "what artificial intelligence could not do was spot most forms of disinformation and misinformation..." The end result... In 2019, other than the government of Vladimir Putin, Warren Buffett was the biggest funder of Sputnik News, the Russian disinformation website controlled by the Kremlin... Geico, the giant American insurance company and subsidiary of Buffett's Berkshire Hathaway, was the leading advertiser on the American version of Sputnik News' global website network... No one at Geico or its advertising agency had any idea its ads would appear on Sputnik, let alone what anti-American content would be displayed alongside the ads. How could they? Which person or army of people at Geico or its agency could have read 44,000 websites? Geico's ads had been placed through a programmatic advertising system that was invented in the late 1990s as the internet developed. It exploded beginning in the mid 2000s and is now the overwhelmingly dominant advertising medium. Programmatic algorithms, not people, decide where to place most of the ads we now see on websites, social media platforms, mobile devices, streaming television, and increasingly hear on podcasts... If Geico's advertising campaign were typical of programmatic campaigns for broad-based consumer products and services, each of its ads would have been placed on an average of 44,000 websites, according to a study done for the leading trade association of big-brand advertisers. Geico is hardly the only rock-solid American brand to be funding the Russians. During the same period that the insurance company's ads appeared on Sputnik News, 196 other programmatic advertisers bought ads on the website, including Best Buy, E-Trade, and Progressive insurance. Sputnik News' sister propaganda outlet, RT.com (it was once called Russia Today until someone in Moscow decided to camouflage its parentage), raked in ad revenue from Walmart, Amazon, PayPal, and Kroger, among others... Almost all advertising online — and even much of it on television (through streaming TV), or on podcasts, radio, mobile devices, and electronic billboards — is now done programmatically, which means the machine, not a planner, makes those placement decisions. Unless the advertiser uses special tools, such as what are called exclusion or inclusion lists, the publishers and content around which the ad appears, and which the ad is financing, are no longer part of the decision. "What I kept hearing as the professionals explained it to me was that the process is like a stock exchange, except that the buyer doesn't know what stock he is buying... the advertiser and its ad agency have no idea where among thousands of websites its ad will appear."

Read more of this story at Slashdot.

Mistral Releases Codestral, Its First Generative AI Model For Code

Mistral, the French AI startup backed by Microsoft and valued at $6 billion, has released its first generative AI model for coding, dubbed Codestral. From a report: Codestral, like other code-generating models, is designed to help developers write and interact with code. It was trained on over 80 programming languages, including Python, Java, C++ and JavaScript, explains Mistral in a blog post. Codestral can complete coding functions, write tests and "fill in" partial code, as well as answer questions about a codebase in English. Mistral describes the model as "open," but that's up for debate. The startup's license prohibits the use of Codestral and its outputs for any commercial activities. There's a carve-out for "development," but even that has caveats: the license goes on to explicitly ban "any internal usage by employees in the context of the company's business activities." The reason could be that Codestral was trained partly on copyrighted content. Codestral might not be worth the trouble, in any case. At 22 billion parameters, the model requires a beefy PC in order to run.

Read more of this story at Slashdot.

❌