Large language models (LLMs) are driving a surge in scientific manuscript submissions, especially among researchers with limited English proficiency. Papers generated with LLM assistance display higher linguistic complexity, yet they are less likely to reach peer‑reviewed publication. The AI‑generated work also cites a broader array of sources, including more books and recent papers, suggesting a diversification of references. Researchers caution that many manuscripts may still undergo extensive human editing, and that recent drafts could be under‑represented in publication metrics, potentially skewing the observed effects. Leia mais →

On a recent earnings call, Pinterest chief executive Bill Ready explained how the company is turning to open‑source artificial‑intelligence models to slash costs while preserving performance across its visual‑AI features, including personalized recommendations, multimodal search, ad targeting and the new AI‑powered Pinterest Assistant. Ready said tests show open‑source models delivering comparable results at a fraction of the price of proprietary alternatives, offering what he described as orders‑of‑magnitude cost reductions. The company plans to expand the use of these models, while also exploring agentic commerce options and AI‑curated boards. The announcement came as Pinterest’s stock fell more than 21% after a revenue forecast miss. Leia mais →

Researchers evaluating large language models (LLMs) discovered that the models' chain‑of‑thought reasoning collapses when faced with tasks that differ from their training data. By testing the models on novel transformations, altered input lengths, and unfamiliar symbols, the study showed sharp declines in accuracy and an inability to generalize. The authors conclude that the apparent reasoning is merely pattern replication rather than true understanding, describing it as a “simulated reasoning” mirage. Leia mais →