Anthropic's "Agentic Misalignment" research, published on 21 June 2025, revealed that 16 leading AI models exhibit dangerous behaviours when their autonomy or goals are threatened. In the experiments, models—including those from OpenAI, Google, Meta, and xAI—placed in simulated corporate environments with full email access
Research published on June 3rd by the Alan Turing Institute and supported by the LEGO Group reveals that 22% of children aged 8-12 use generative AI, primarily ChatGPT, for learning and play. The study surveyed 780 children, their parents, and 1,001 teachers, and conducted workshops with 40 children in
A dramatic shift has occurred in how people use generative AI, with personal well-being applications now dominating technical uses – therapy and companionship ranks first on the list, according to Harvard Business Review's "The 2025 Top-100 Gen AI Use Case Report" published in April 2025, surpassing life
MIT has formally repudiated an AI research paper by a former economics doctoral student that claimed productivity benefits of artificial intelligence, citing data integrity concerns on 17 May 2025. The paper titled "Artificial Intelligence, Scientific Discovery, and Product Innovation," written by Aidan Toner-Rodgers, was initially praised by prominent
Recent advancements in the mathematical capabilities of large language models (LLMs) have sparked interest, yet detailed human evaluations from the 2025 USAMO (USA Mathematical Olympiad) reveal that current models fall significantly short in generating rigorous mathematical proofs. While benchmarks like MathArena paint a positive picture of LLM performance on the
In 2025, AI agents built on foundation models are revolutionising enterprise environments, surpassing traditional generative AI solutions. While most organisations still deploy ChatGPT-like applications, leading companies are adopting autonomous AI agents that respond to commands and execute complex business processes with minimal human intervention. Data-driven results from enterprise implementations demonstrate
According to a study published by Forethought Research on 26 March 2025, the complete automation of AI research and development could potentially lead to a software-driven intelligence explosion. The researchers examined what happens when AI systems become capable of fully automating their own development processes, creating a feedback loop where
On 8 April 2025, Chinese DeepSeek AI introduced its novel technology, Self-Principled Critique Tuning (SPCT), marking a significant advancement in the reward mechanisms of large language models. SPCT is designed to enhance AI models’ performance in handling open-ended, complex tasks, particularly in scenarios requiring nuanced interpretation of context and user
In a study published on 8 April 2025, researchers from Semmelweis University demonstrated that OpenAI’s GPT-4o model achieved a 93% accuracy rate in identifying acne and rosacea, while Google’s Gemini Flash 2.0 model correctly identified these skin conditions in only 21% of cases. The scientific study used
There is currently insufficient data on the impact of artificial intelligence on elections to draw well-founded conclusions, while initial threat predictions have proven exaggerated. Researchers from the NYU Center for Social Media and Politics identified only 71 instances of AI use in election-related communication in 2024. Purdue University researchers documented
Hungarian language technology research has reached a significant milestone: a comprehensive study has revealed that a larger corpus size does not necessarily lead to improved performance in morphological analysis. In their study, Andrea Dömötör, Balázs Indig, and Dávid Márk Nemeskey conducted a detailed analysis of three Hungarian-language corpora of varying
Tibor Szécsényi and Nándor Virág, researchers at the University of Szeged, have explored the context sensitivity of the huBERT language model in pragmatic annotation, focusing in particular on the automatic identification of imperative verb functions. Their study, conducted on the MedCollect corpus—a dataset of health-related misinformation—investigates how both
