Uno studio pubblicato su Nature Health ha analizzato quasi 900.000 risposte generate da cinque modelli di intelligenza artificiale, scoprendo una forte tendenza a privilegiare pochi prodotti farmaceut...

Since ChatGPT3′s chatbot interface appeared in November 2022, AI cheating has re-incarnated a challenge for academia. It is accentuated in non-invigilated contexts during term as projects, essays a...

Mistral AI SAS today announced the availability of the open weights of its frontier model Mistral Large 4, nicknamed “Le Chonk,” the company’s most capable large language model to date, initially...

Background: There is evidence of rapid adoption of large language models (LLMs) in undergraduate medical education, with medical students using them to support their learning. This raises concerns about the accuracy of LLM-generated information that students are learning from. Retrieval-augmented generation (RAG), which grounds an LLM in curated materials, may be a strategy to improve the accuracy of this information. Objective: This study explored medical students’ perceptions after using an online tool, NotebookLM, that uses RAG. Methods: A cohort of 284 second-year medical students at Bristol Medical School were encouraged to use NotebookLM over a period of 4 months (March 2025 to June 2025) to help them learn course content. Following this, volunteers were recruited to share their perspectives through interviews and a survey that was open for 4 months (June 2025 to September 2025). A total of 9 individuals participated: 2 through interviews and 7 through the survey. Qualitative data were analyzed using reflexive thematic analysis and supplemented by Likert scale data. Results: NotebookLM was perceived as more reliable than standard LLM responses by interview participants, mainly due to the transparency of and control over information sources. Reliability ratings from survey participants were more mixed, but none rated unreliable. Students described it as an accessible, on-demand learning resource, which they perceived as improving efficiency and supporting self-directed study. They reported continued use of strategies to cross-check generated information. Participants demonstrated some awareness of ethical concerns, but it was unclear whether this altered their practice. Conclusions: In this small, exploratory study, students described NotebookLM as improving trust in LLM outputs while maintaining their flexibility. The findings suggest that RAG may be a strategy to support students who use LLMs. However, further research is needed to formally evaluate RAG vs standard LLMs, including the impact of RAG on accuracy, curricular alignment, and the ability of students to meet learning outcomes.

Inception introduced Mercury 3 on October 19, 2026, describing the model as the most capable diffusion-based large language model released to date, with launch pricing of $0.10 per million input...

The stubbornly human business of knowing what's funny.

Background: Large language models (LLMs) could assist in streamlining health care instrument validation. This is particularly relevant as traditional instrument development is time-intensive and resource-intensive. Currently, there is a significant gap in specifically developed and validated trust instruments for pediatric emergency and surgical contexts. Such tools are key because trust is foundational to the patient family–physician relationship and is associated with improved care-seeking and treatment adherence. Objective: This study aimed (1) to develop new trust instruments and conduct their preliminary content validation confirming item relevance through a Synthetic Instrument Validation approach integrating human and LLM capabilities and (2) to evaluate appropriate use cases for LLMs in psychometric assessment. Methods: Two trust instruments were developed (patient family and physician versions). In phase 1, instruments underwent a 2-stage content validation process using parallel synthetic and human expert panels (5 patient family members, 3 physician synthetic experts, and 7 unique human experts). Synthetic panels consisted of 3 persona-prompted LLMs (Claude Sonnet 4, GPT-5, and Grok 4), with human panels serving as comparators. The scale–content validity index (S-CVI) threshold was set at ≥0.80, percent agreement on item relevancy was used to compare panels. In phase 2, LLM performance in quantitative research tasks was evaluated. The patient family instrument underwent Flesch-Kincaid assessment; both instruments underwent cosine similarity analyses using algorithmic and LLM-based methods. Results: In phase 1, percent agreement between panels for the patient family and physician instruments was 90.9% and 88.9% for dimensional validation and 90.9% and 100% for contextual validation. Fleiss κ for within-panel agreement ranged from poor to almost perfect; 0.20 (95% CI -0.23 to 0.63) to 1.00 (95% CI 1.00-1.00). All instrument subsections met or exceeded S-CVI thresholds across both rounds and panels. Phase 2 Flesch-Kincaid metrics converged across all 3 methods to target benchmark (mean Grade Level 8.1, SD 1.1; mean Reading Ease 60.1, SD 5.6), while reference readability was near target, showing convergence across implementations of the same formula. In contrast, cosine similarity analyses revealed significant LLM performance limitations; cosine similarity analyses were, therefore, abandoned in favor of the algorithmic method alone, which achieved a maximum cosine similarity of 0.83, indicating acceptable item distinctiveness overall. Conclusions: Persona-prompted LLMs effectively assisted preliminary content validation and may help reduce timelines from months to weeks, but showed limitations in quantitative computations. This suggests that, in this study, persona-prompted LLMs performed better in subjective rating tasks than in deterministic cosine similarity computations. These findings help illustrate task-dependent boundaries for LLM integration in psychometric research, necessitating selective human-LLM collaboration. While these instruments achieved preliminary content validity, they require full psychometric testing before they can be considered validated for clinical use. Synthetic Instrument Validation offers a feasible approach to preliminary health care instrument development while preserving key elements of content validation.

Meta has launched Muse Wonder and Muse Wonder Full, its first frontier-class AI models from Meta Superintelligence Labs, featuring a 1 million token context window and competitive pricing against OpenAI and Anthropic offerings.

Código Abierto · Episode

Se argumenta que la representación del conocimiento que parecía perdida en la IA, parece retornar con fuerza en aplicaciones como los agentes.