Mexico Journalists, Persistent American Poverty, Understanding LLMs, More: Wednesday ResearchBuzz, May 10, 2023


Rest of World: The online repository that keeps the work of Mexico’s murdered journalists alive. “In the past two decades, over 150 journalists have been killed in the country. Defensores de la Democracia is building a living online archive to preserve their work.”

Census Bureau: Census Bureau Releases New Report About Persistent Poverty at County and Census-Tract Level. “The report examines county and subcounty geographies (specifically, census tracts), identifying additional populations that may benefit from targeted intervention. It explores census tracts over the same 30-year period (1989 to 2015-2019) as counties allowing for a more direct comparison.”

TechCrunch: OpenAI’s new tool attempts to explain language models’ behaviors. “It’s often said that large language models (LLMs) along the lines of OpenAI’s ChatGPT are a black box, and certainly, there’s some truth to that. Even for data scientists, it’s difficult to know why, always, a model responds in the way it does, like inventing facts out of whole cloth. In an effort to peel back the layers of LLMs, OpenAI is developing a tool to automatically identify which parts of an LLM are responsible for which of its behaviors. The engineers behind it stress that it’s in the early stages, but the code to run it is available in open source on GitHub as of this morning.”

Niagara Frontier Publications: ABC News Digital debuts multimedia project ‘Buffalo: Healing from Hate – One Year Later’. “ABC News announced ABC News Digital’s ‘Buffalo: Healing from Hate – One Year Later,’ a multimedia project chronicling the one-year mark of the May 14, 2022, mass shooting that claimed the lives of 10 people at the Tops supermarket in Buffalo.”


The 19th: ‘The world’s largest Black group chat’: Behind the mission to preserve Black Twitter. “[Meredith] Clark is part of Archiving The Black Web, a group of digital archivists seeking to preserve the stories of Black people and extend existing archival practices to the digital sphere. This group and others hope to document not just the content created on the platform but how Black people use it for communication and community.”

Washington Post: Fake sign language is spreading on TikTok. Deaf people are worried.. “Anthony Eagle Jr. is big on TikTok. He boasts over 850,000 followers, many of whom love the way he performs sign language renditions of songs. There’s just one problem — the sign language is sometimes wrong. When Eagle, 39, of Winston-Salem, N.C., signs the song, ‘Love the Way You Lie,’ his rendition is riddled with mistakes, like signing the word ‘lie’ with two hands in the wrong position. To a deaf person who uses sign language, it looks like gibberish.”


BBC: Briton pleads guilty in US to 2020 Twitter hack. “A British national extradited to the US last month has pleaded guilty in New York to a role in one of the biggest hacks in social media history. The July 2020 Twitter hack affected over 130 accounts including those of Barack Obama and Joe Biden. Joseph James O’Connor, 23, known as PlugwalkJoe, pleaded guilty to hacking charges carrying a total maximum sentence of over 70 years in prison.”

Ars Technica: Feds seize 13 more DDoS-for-hire platforms in ongoing international crackdown. “The US Justice Department has sized the domains of 13 DDoS-for-hire services as part of an ongoing initiative for combatting the Internet menace. The providers of these illicit services platforms describe them as ‘booter’ or ‘stressor’ services that allow site admins to test the robustness and stability of their infrastructure. Almost, if not all, are patronized by people out to exact revenge on sites they don’t like or to further extortion, bribes, or other forms of graft.”

The Guardian: Why is Google stonewalling regulation in Brazil?. “Newspaper Folha de S Paulo reported that Google’s strategy included sending emails to YouTubers saying there would be less money to invest in their channels and asking them to talk to their Congress. The tech giant also fumbled with search results, prominently showing its own blog post and other articles that were critical of the bill, according to a study by the Federal University of Rio de Janeiro.”


CTV News: Pinterest identifies 1,403% more child abuse material in 2022, but majority of reports come from Facebook. “Major social media sites and digital platforms reported a nine per cent increase in suspected child sexual abuse material in 2022, with 85.5 per cent of all 31.8 million reports coming from Meta platforms Facebook, Instagram and WhatsApp.”

Toronto Star: Canada set to name foreign labs, universities that pose risk to national security. “Ottawa is in ‘advanced stages’ of drafting a list of entities that pose a risk to national security, and top universities are prepared to avoid working with these entities despite what could be a loss of $100 million or more in annual research funding from foreign partners.”

The Washington Post: Google promised to delete sensitive data. It logged my abortion clinic visit.. “When the Supreme Court overturned Roe v. Wade, privacy advocates, including me, raised an alarm that data from smartphones could be used to help prosecute abortions. Google offered a partial solution: It would proactively delete its trove of location data when people visited ‘particularly personal’ places, including abortion clinics, hospitals and shelters. Nearly a year later, my investigation reveals Google isn’t doing that in any consistent way. And its response to me shows it isn’t taking accountability.” Good morning, Internet…

Do you like ResearchBuzz? Does it help you out? Please consider supporting it on Patreon. Not interested in commitment? Perhaps you’d buy me an iced tea. Check out Search Gizmos when you have a minute.

Categories: morningbuzz

Leave a Reply