No news signals yet.
IBM reported a disappointing quarter with revenue of $17.2 billion, missing Wall Street's expectations. CEO Arvind Krishna attributed the 42% drop in mainframe sales to customers prioritizing other ha…
Synthesia has launched Roleplay Sessions, an AI-powered training platform where employees practice high-stakes conversations with an avatar. The British startup's new product aims to help enterprises…
Researchers studied Large Language Models' persuasive potential using Jürgen Habermas' Theory of Communicative Action. They found that LLMs effectively convey illocutionary intent, often more so than…
Researchers identify Safety Drift and Operational Hallucination in large language models, where initial alignment degrades over time, leading to constraint-violating actions. These phenomena are obser…
Researchers introduce Interpretive Knowledge Nodes and Citation Compression Layers to analyze citation pathways in the AI era. AI transforms the cost structure of producing citable knowledge intermedi…
Researchers introduce a mathematical framework to study AI value alignment with evolving social norms, highlighting risks like value lock-in and normative mode collapse. The framework uses social phys…
Researchers analyzed 2,611 Spanish mobile apps using large language models, finding discrepancies between declared and observed privacy practices, especially in public-sector apps. The study reveals l…
Researchers address gaps in the concept of loss of control in AI, establishing a definition of control based on setting and getting goals, and discussing how control can be lost, including through AI…
Researchers propose MR-ConceptGCN, a multi-relational graph convolutional network approach for sequential learner modeling, combining personal knowledge graphs and pre-trained language models. The met…
Researchers propose a Two TimeScale Reinforcement Learning framework for delivery drone-based participatory sensing in dynamic environments. The framework, called TSRL, separates decision-making into…
Researchers analyzed 18,797 work tasks in 124 economies, finding that feasible automation depends on task content and country-level conditions. The exposed share of tasks ranges from 3.3% to 61.6%, va…
Researchers compared TF-IDF + Logistic Regression and fine-tuned DistilBERT for phishing email detection, finding both exceeded 98% accuracy on clean data but dropped to 64.00% and 63.64% under advers…
Researchers argue that modern AI systems face hidden safety-critical challenges, including quieter failures that are plausible and distributed across components. A proposed five-layer framework identi…
Researchers evaluated 41,331 summaries from 13,777 articles, finding AI-powered browsers like Google Chrome and Microsoft Edge to be broadly accurate and effective at reducing political bias and negat…
Researchers compared GPT-generated surveys to human-designed ones across three social domains. The GPT-generated surveys captured dominant attitudinal divisions, but showed differences in belief struc…
African governments are prioritizing AI opportunities for economic transformation, but devote little attention to AI safety, according to an analysis of national AI strategies and public statements. T…
Researchers introduce digital statecraft as a concept to govern the digital substrate of states, comprising data, algorithms, and infrastructure. The concept operates on two dimensions, requiring tech…
Researchers analyzed NFL tweets from 2011-2014, finding sentiment for teams is positive before games, drops at kickoff, and diverges after halftime, with winning teams' sentiment increasing and losing…
Anthropic and Physical Intelligence held acquisition talks this spring, according to The Information, despite Physical Intelligence's CEO denying the rumor. Physical Intelligence, co-founded by Lachy…
Dimension Capital, a New York firm, has announced an $800 million fund, 60% larger than its previous $500 million fund. Founded by Zavian Dar, Adam Goulburn, and Nan Li, the firm invests in companies…
Meta is piloting StoryKit, an AI storytelling app, in select countries to create personalized children's stories with custom characters and lessons. The app uses AI safety filters and is available to…
Gritt, a startup founded by Puneet Puri and Vishal Dugar, has exited stealth with $34 million in funding to develop robots that build solar plants. The company uses AI models to control off-the-shelf…
A study analyzed 107 papers, 114 Reddit posts, and 300 comments to understand students' perceptions of peer grading. Students view peer grading as both beneficial and problematic, citing benefits like…
Researchers propose a secure liquid democracy mechanism to mitigate voting power imbalances. The system uses decentralized timed-release encryption to seal delegation choices and ranked backup delegat…
Researchers introduce STRATA, a model that infers race and ethnicity from names and geography, achieving 89.2% accuracy. STRATA reduces the White False Positive Rate from 41.8% to 17.8%, outperforming…
Researchers redesigned a university-level Introduction to Artificial Intelligence course, incorporating interactive simulations and collaborative labs. Students in the redesigned course reported highe…
A study of 410 German companies found that AI tools enhance HR analytics capabilities, mainly serving efficiency and rationalising goals. Organisational transformation factors, such as digital infrast…
Researchers studied guided large language model use in an undergraduate Probability and Statistics course, finding that guided access was associated with clearer learning-oriented interaction patterns…
Researchers propose a conceptualisation of systemic risks of AI, focusing on complex externalities and collective action problems, including feedback dynamics and market concentration. This approach c…
Researchers evaluated Portugal's 9B language model, AMALIA, as a scientific instrument, finding it agrees with human coders but only attributes half of its coding performance to underlying theory. A l…