There is now a solid body of evidence showing that internet availability is causing a variety of outcomes that adversely affect democracy
Algorithms are major culprits becase of their focus on maximizing user engagement with little regard for the quality of content. www.science.org/doi/10.1126/...
Check out this new short series on trust and safety in gaming!
It was great to put this together with @sbradshaw.bsky.social @dwj.eurosky.social Mariana Olaizola Rosenblat, and @kateruane.bsky.social.
In the "you can't make this up" category of LLM interactions...
TL;DR Google Gemini claims that LLMs "often give themselves a 95%+ confidence score even when they are hallucinating" - which was itself a "hallucinated" claim - then later assigns itself 0.95 confidence in that hallucinated claim. 🧵
There is so much debate about what AI will do to us in the future.
But @proofnews.bsky.social is reporting about what AI is doing to us right now. This week we dove deep into an AI robot named Moxi that has been deployed to more than 25 hospital systems across the U.S.
www.proofnews.org/moxi/
Data reviewed by The Post shows that President Trump has selected fewer women to serve in these senior roles than during the same period of his first term.
There also has been a historic amount of turnover among women in Trump’s Cabinet.
I am begging people to realize that police do not use this technology is not a "jumping off point" or an "investigative lead"--the computer spits out a name and police arrest that person without follow up. How many people have to spend months in jail "by accident" before we consider the danger?
AI isn't making translation easier, it's actually making it more cognitively taxing for translators to review error-laden text, requiring more work in less time, and making them less money, reports Sarah O'Connor for the Financial Times www.ft.com/content/75bc...
safety evaluation research the past year or two:
"we use llms to ideate & define “harm”, then use llms to generate prompts about it, which we then feed to llms to generate responses, we then use an llm-as-a-judge to classify the responses, we then use an llm to evaluate the accuracy…"
Jay Van Bavel, PhD
Spinning in circles and crying for help, Moxi robots became another charge to care for, nurses said.
I think I speak for everyone when I ask how this will be used to sell advertisements.
Dr. Casey Fiesler
CDT is embarking on new research to understand how users in the Global Majority experience genAI systems, specifically popular AI chatbots, and how companies develop and enforce their safety policies to take their languages and use cases into context.
cdt.org/insights/ass...
Julia Angwin
A San Diego man spent one month in jail for a crime he didn't commit, after San Diego Police misinterpreted a hit from a license plate reader.
Toronto-based start-up Aurmada is designing AI-enhanced clothing that it hopes will help those wearing it monitor everything from their gait to how close someone is standing, part of a budding wearabl...
AI systems do not work consistently and reliably across languages, as CDT has previously documented. Although companies, researchers, and civil society are working toward filling necessary technical a...
A Post review shows the president built a more male-dominated Cabinet this term, and every departure has been a woman — with a man chosen to replace each of them.
From the “Polybius” urban legend to today’s debates over online harms, gaming has long been the subject of public anxiety. A new CDT essay series, Panic at the Arcade?, explores what games can teach us about trust & safety, and why evidence-based approaches matter. cdt.org/insights/pan...
