{"ok":true,"nota":{"id":"3553008fde3ceeec","slug":"cinco-chatbots-de-ia-erraron-en-casos-de-urgencias-y-uno-solo-fue-apen-355300","titulo":"Cinco chatbots de IA erraron en casos de urgencias y uno solo fue apenas mejor","resumen":"Investigadores compararon cinco modelos de lenguaje con 17 casos ficticios de guardia, evaluados a ciegas por dos médicos de urgencias. Todos nombraron algún diagnóstico correcto, pero las respuestas variaron en triage, calidad y potencial de daño.","fuente":"medrxiv","fuenteNombre":"medRxiv · preprints","color":"#1f7a4d","autor":"Johal, G., Liang, L.","url":"https://www.medrxiv.org/content/10.64898/2026.10.21.26366076v1?rss=1","imagen":"/hoy/portada/3553008fde3ceeec","tema":"tecnologia","subtemas":["inteligencia artificial","chatbots","medicina de urgencias","triage","seguridad del paciente"],"entidades":["Claude","GPT","Grok","Gemini","Kimi"],"formato":"dato","entretenimiento":false,"pubTs":"2026-10-23T00:00:00.000Z","indexTs":"2026-10-23T18:12:31.737Z","hash":"506838a254858066db47548897f826640c2268ebe5ad0c6d783e201b37b827ba","tsa":{"time":"Oct 23 18:13:30 2026 GMT","provider":"freetsa.org","sobre":"sello"},"stats":{"views":0,"likes":0,"clicks":0,"dwell":0},"tipo":"motor","registro":"hallazgo","porQueImporta":"Cada vez más gente le consulta a un chatbot qué hacer cuando un familiar se siente mal de noche, y el estudio muestra que todos los modelos acertaron en algo pero fallaron en otra parte. Para alguien de acá que está decidiendo si ir a la guardia o esperar, una respuesta segura puede ser una respuesta equivocada. Los propios autores sugieren que la herramienta sirve para orientarse, no para decidir sola. Confiar en la IA para una urgencia es jugar con algo que no se recupera.","quienFirma":"Equipo de investigación en medicina de emergencias (institución no especificada en el resumen)","confianza":"preprint","consulta":{"pregunta":"¿Le preguntarías a un chatbot si tenés que ir a la guardia?"}},"sello":{"hash":"506838a254858066db47548897f826640c2268ebe5ad0c6d783e201b37b827ba","sello":"2a45bcd2fb103236485b7db375d1b94ff28b28c48c10d83e3c51fc5c9d5a1b01","beacon":{"drand":{"cadena":"8990e7a9aaed2ffed73dbd7092123d6f289930540d7651336225dc172e51b2ce","ronda":6578272,"aleatorio":"e504aca3fcd7243fddbec8b88f1c51dc96cb1f827ccda23c013f3b302f3c69ab","tiempo":"2026-10-23T18:13:00.000Z"},"nist":{"cadena":2,"pulso":1996329,"salida":"880EE1B503674E57D0E81D7CEC0FFDBB3A3BB5CD5A62E05BB575BBCB9E2DB29F611799CDCF1B120D62E8DF3A39AE0D1B9935A0A198FC708C5974107E51CFF0C4","tiempo":"2026-10-23T18:12:00.000Z"},"noAntes":"2026-10-23T18:13:00.000Z"},"verificacion":{"drand":true,"nist":true},"tsa":{"time":"Oct 23 18:13:30 2026 GMT","provider":"freetsa.org","sobre":"sello"},"original":{"titulo":"Diagnosis, triage and response quality of five large language models on emergency medicine vignettes: a blinded comparison","descripcion":"Background: Emergency care requires appropriate triage and management as well as diagnosis. We compared five large language models on these tasks using the same clinical vignettes. Methods: Seventeen synthetic emergency department vignettes were submitted to Claude Fable 5, GPT-5.6 Sol, Grok 4.5, Gemini 3.6 Flash and Kimi K3 through consumer interfaces on 11-12 August 2026. Two emergency physicians independently assessed the 85 responses with model identities concealed. Outcomes were agreement with reference diagnoses and Canadian Triage and Acuity Scale (CTAS) levels, appropriateness and overall quality (1-5 scales), potential harm and preferred response. Results: Every model correctly name","pubDate":"2026-10-23","url":"https://www.medrxiv.org/content/10.64898/2026.10.21.26366076v1?rss=1"},"indexado":"2026-10-23T18:12:31.737Z"}}