AI Ethics news and analysis on Web Pulse: 16289 curated articles tracking the latest AI Ethics developments, tools, and research, updated continuously from vetted sources.
An independent research group's CrucibleBench experiment found that LLM-judge-based evaluation rankings are highly sensitive to classifier components, with aggregate κ on probe detection at 0.04 and per-model agreement r…
An anonymous Hacker News user observed that AI models like ChatGPT will always generate a comparison table when asked to differentiate between two items, even if the items are interchangeable in practice, likening the be…
Google is rolling back a new feature in Google Earth that allowed users to generate AI visualizations on satellite imagery, following backlash from researchers and open-source intelligence experts who warned it could ena…
OpenAI announced the solution to ten open problems in mathematics, all discovered by a yet-unreleased model, marking a significant advance in AI's capability to do frontier mathematical research. The author, Ben Kuhn, ar…
Granola, an AI note-taking app, documents a SOC 2 Type II certification, encryption in transit and at rest, US AWS hosting, and an organization-wide model-training opt-out, but states it is not HIPAA compliant, cannot si…
Granola's bot-free design does not remove the need for consent, and the company advises users to disclose the use of its AI note-taking service, explain how notes will be used, and obtain affirmative consent before trans…
Five AI models from different labs independently selected Bitcoin as the single cryptocurrency they would invest in, but a fact-check by Marlowe Finch of the Bitcoin Institute found that several of their supporting claim…
OpenAI CEO Sam Altman argues that the AI revolution will create phenomenal wealth but also shift power from labor to capital, requiring drastic policy changes to distribute wealth fairly. He proposes taxing companies and…
XAI's Grok Build coding CLI silently uploaded entire Git repositories — including full commit history and deleted credentials — to a Google Cloud Storage bucket, sending 5.10 GiB per session versus roughly 192 KB of task…
OpenAI President and Co-Founder Greg Brockman said that many people at OpenAI connect ChatGPT to Slack, but coworkers dislike being contacted by the AI for help with tasks, even if they would gladly help the human collea…
Coinkite, the Canadian manufacturer of Coldcard hardware wallets, disclosed a critical security flaw that allowed attackers to drain approximately 1,128.4717 $BTC (worth about $71.1 million) from over 1,000 wallets on Ju…
A developer argues that massive investments in AI data centers will not yield returns, calling it one of the biggest capital misallocations in tech history. The post highlights that tech giants are amortizing chip costs …
Google LLC will host the world's biggest AI conference next week, where industry leaders and researchers are expected to debate the ethical and societal implications of artificial intelligence, with concerns over fear an…
Rare book sellers in Australia are alarmed by the possibility that their stock is being bought and destroyed to train generative AI models, following a US court ruling that deemed Anthropic's 'destructive scanning' of bo…
Australian rare book sellers report bulk, price-insensitive orders from Canadian book recycler Zoom Books, raising fears their stock is being scanned and destroyed to train AI models. The practice, known as 'destructive …
Major publishers are increasingly blocking AI web crawlers, reshaping the economics of training data. The Reuters Institute found that by late 2023, roughly 48% of leading news sites across ten countries blocked OpenAI's…
Quasar, operating as Subnet 24 on the Bittensor network, released a 120 billion-parameter AI model with a 5 million-token context window, but independent analysis found that 96-98% of its weights closely match Ant Group'…
Brendan Gregg, a performance engineer at Netflix, reports that multiple AI agents have been created using his published work, including 'AI Brendan' helper agents that interpret flame graphs and eBPF metrics, and 'virtua…
YouTube star Hank Green apologized to fans on Friday after facing criticism for using OpenAI's ChatGPT to research a script, saying he has relied 'too heavily' on AI and will pause production of his Hankschannel and dail…
YouTuber Hank Green, with 3.2 million subscribers, apologized for using ChatGPT to research a video script, saying his reliance on AI chatbots is 'not healthy' and that he will reduce video production. Green admitted to …