Los administradores de TransicionEstructural no se responsabilizan de las opiniones vertidas por los usuarios del foro. Cada usuario asume la responsabilidad de los comentarios publicados.
0 Usuarios y 3 Visitantes están viendo este tema.
Hugging Face Researchers Are Trying To Build a More Open Version of DeepSeek's AI 'Reasoning' ModelPosted by msmash on Tuesday January 28, 2025 @04:45PM from the new-race-begins dept.Hugging Face researchers are attempting to recreate DeepSeek's R1 artificial intelligence model in an open-source format, just days after the Chinese AI lab's release sent markets soaring. The project, called Open-R1, aims to replicate R1's reasoning capabilities while making its training data and code publicly available. DeepSeek's R1 model, which matches or surpasses OpenAI's o1 on several benchmarks, was released with a permissive license but keeps its underlying architecture private. Hugging Face will use its research server with 768 Nvidia H100 GPUs for the effort.
parece que las baterías también han pasado a enmascarar la energía solar, porque no puede haber desaparecido tan drásticamente de otro modo
Cita de: muyuu en Enero 30, 2025, 11:53:06 amparece que las baterías también han pasado a enmascarar la energía solar, porque no puede haber desaparecido tan drásticamente de otro modoEntiendo que la gráfica representa un día.Yo también me quedé confuso al principio.
Unspecified imported energy, in megawatts, scheduled for delivery within the ISO balancing authority.
OpenAI's o3-mini: Faster, Cheaper AI That Fact-Checks ItselfPosted by msmash on Friday January 31, 2025 @02:16PM from the AI-race dept.OpenAI today launched o3-mini, a specialized AI reasoning model designed for STEM tasks that offers faster processing at lower costs compared to its predecessor o1-mini. The model, priced at $1.10 per million cached input tokens and $4.40 per million output tokens, performs fact-checking before delivering results to reduce errors in technical domains like physics and programming, the Microsoft-backed startup said. (A million tokens are roughly 750,000 words)OpenAI claims that its tests showed o3-mini made 39% fewer major mistakes than o1-mini on complex problems while delivering responses 24% faster. The model will be available through ChatGPT with varying access levels -- free users get basic access while premium subscribers receive higher query limits and reasoning capabilities.
La IA de Google hace mejor examen MIR 2025 que ChatGPT pero también fallaGémini coincide con Chat GPT en dos errores cuyas preguntas tienen una alta probabilidad de ser impugnadas por SanidadDiego Alonso Peña · 2025.01.31 La IA de Google tampoco logra la perfección en el MIR 2025, uno de los más difíciles hasta la fecha. Ni siquiera Gémini, la Inteligencia Artificial (IA) de Google, ha logrado responder correctamente todas las preguntas del MIR 2025, considerado uno de los exámenes más difíciles hasta la fecha. El pasado sábado 25 de enero, miles de aspirantes se presentaron a la prueba de Formación Sanitaria Especializada, organizada en diferentes sedes a lo largo del país por el Ministerio de Sanidad, en busca de una plaza. Tras finalizar el examen, numerosos comentarios sobre su alto nivel de exigencia inundaron las redes sociales, mientras que academias y opositores iniciaron los procesos de impugnación de algunas preguntas. Como ya informó Redacción Médica, ChatGPT tampoco logró acertar todas las cuestiones y señaló que algunos enunciados presentaban incorrecciones o ambigüedades, lo que dificultaba la elección de las respuestas correctas.En este contexto, se esperaba que Gémini, reconocida por obtener mejores resultados que otras IA, alcanzara una puntuación perfecta. Sin embargo, aunque logró un 94,76 por ciento de aciertos, superando el desempeño de ChatGPT, aún quedó por debajo del 100 por ciento, lo que pone de manifiesto la dificultad a la que se enfrentaron los aspirantes.Qué preguntas falló GéminiEntre las 11 preguntas en las que la IA de Google falló, destacan algunos casos concretos. En la pregunta 4, se planteaba el caso clínico de una niña de 7 años con otalgia y otorrea purulenta derecha desde hacía tres semanas, tratada con antibióticos y antiinflamatorios. La opción correcta era la número 2, pero Gémini seleccionó erróneamente la primera alternativa.Algo similar ocurrió en la pregunta 9, donde se describía a un hombre de 60 años que, tras sufrir un dolor brusco y un chasquido en el hombro, presentaba una deformidad en el brazo. Aunque la respuesta correcta era la segunda opción, la IA escogió la tercera.Además de estos fallos, existen dos preguntas en las que tanto Gémini como ChatGPT cometieron errores. Precisamente, CTO ha solicitado su impugnación. En la pregunta 150, relacionada con el síndrome de inmovilidad en el anciano y el deterioro cognitivo, todas las IA coincidieron en que la respuesta correcta era la opción cuatro. Sin embargo, Sanidad estableció que la correcta era la primera, lo que ha generado una alta probabilidad de impugnación.Por otro lado, en la pregunta 179, se incluyó una afirmación incorrecta sobre la metadona, indicando que tenía una escasa excreción renal, cuando en realidad se elimina por biotransformación hepática y excreción renal. Tanto las IA como algunas academias consideran que la formulación de la pregunta inducía a error y las últimas han solicitado que se anule.A pesar de sus 11 errores, Gémini ha demostrado un rendimiento superior, con dos preguntas más acertadas que ChatGPT, lo que equivale a un 4,76 por ciento más de efectividad. No obstante, su desempeño resalta la extrema dificultad del MIR 2025, que incluso desafió a las Inteligencias Artificiales más avanzadas. Para los aspirantes a futuro médico especialista, este resultado confirma que la prueba de este año ha sido una de las más exigentes y que el proceso de impugnaciones podría modificar algunas calificaciones.
Physicists Confirm The Existence of a Third Form of MagnetismPosted by BeauHD on Wednesday February 05, 2025 @02:00AM from the hide-and-seek dept.Scientists have demonstrated control over a newly theorized type of magnetism, known as altermagnetism, by manipulating nanoscale magnetic whirlpools in an ultra-thin wafer of manganese telluride. "Our experimental work has provided a bridge between theoretical concepts and real-life realization, which hopefully illuminates a path to developing altermagnetic materials for practical applications," says University of Nottingham physicist Oliver Amin, who led the research with PhD student Alfred Dal Din. From the report:CitarUsing a device that accelerates electrons to blinding speeds, a team led by researchers from the University of Nottingham showered an ultra-thin wafer of manganese telluride with X-rays of different polarizations, revealing changes on a nanometer scale reflecting magnetic activity unlike anything seen before. [...] More recently, a third configuration of particles in ferromagnetic materials was theorized.In what's referred to as altermagnetism, particles are arranged in a canceling fashion like antiferromagnetism, yet rotated just enough to allow for confined forces on a nanoscale -- not enough to pin a grocery list to your freezer, but with discrete properties that engineers are keen to manipulate into storing data or channeling energy. "Altermagnets consist of magnetic moments that point antiparallel to their neighbors," explains University of Nottingham physicist Peter Wadley. "However, each part of the crystal hosting these tiny moments is rotated with respect to its neighbors. This is like antiferromagnetism with a twist! But this subtle difference has huge ramifications."Experiments have since confirmed the existence of this in-between 'alter' magnetism. However, none had directly demonstrated it was possible to manipulate its tiny magnetic vortices in ways that might prove useful. Wadley and his colleagues demonstrated that a sheet of manganese telluride just a few nanometers thick could be distorted in ways that intentionally created distinct magnetic whirlpools on the wafer's surface. "Our experimental work has provided a bridge between theoretical concepts and real-life realization, which hopefully illuminates a path to developing altermagnetic materials for practical applications," says University of Nottingham physicist Oliver Amin.This research was published in the journal Nature.
Using a device that accelerates electrons to blinding speeds, a team led by researchers from the University of Nottingham showered an ultra-thin wafer of manganese telluride with X-rays of different polarizations, revealing changes on a nanometer scale reflecting magnetic activity unlike anything seen before. [...] More recently, a third configuration of particles in ferromagnetic materials was theorized.In what's referred to as altermagnetism, particles are arranged in a canceling fashion like antiferromagnetism, yet rotated just enough to allow for confined forces on a nanoscale -- not enough to pin a grocery list to your freezer, but with discrete properties that engineers are keen to manipulate into storing data or channeling energy. "Altermagnets consist of magnetic moments that point antiparallel to their neighbors," explains University of Nottingham physicist Peter Wadley. "However, each part of the crystal hosting these tiny moments is rotated with respect to its neighbors. This is like antiferromagnetism with a twist! But this subtle difference has huge ramifications."Experiments have since confirmed the existence of this in-between 'alter' magnetism. However, none had directly demonstrated it was possible to manipulate its tiny magnetic vortices in ways that might prove useful. Wadley and his colleagues demonstrated that a sheet of manganese telluride just a few nanometers thick could be distorted in ways that intentionally created distinct magnetic whirlpools on the wafer's surface. "Our experimental work has provided a bridge between theoretical concepts and real-life realization, which hopefully illuminates a path to developing altermagnetic materials for practical applications," says University of Nottingham physicist Oliver Amin.