https://aiexpert.news/en/article/big-tech-capex-hits-725b-as-chip-memory-costs-soarmicrosoft-alone-spends-25b-on- ai|expert en 2026-05-03T22:26:25.274Z Microsoft's $25B Memory Inflation Swallows Capex Gains https://aiexpert.news/pt/article/big-tech-capex-hits-725b-as-chip-memory-costs-soarmicrosoft-alone-spends-25b-on- ai|expert pt-BR 2026-05-03T22:26:25.274Z A Inflação de Memória de $25B da Microsoft Engole Ganhos em Capex https://aiexpert.news/es/article/big-tech-capex-hits-725b-as-chip-memory-costs-soarmicrosoft-alone-spends-25b-on- ai|expert es 2026-05-03T22:26:25.274Z La Inflación de Memoria de $25B de Microsoft Anula las Ganancias en Capex https://aiexpert.news/en/article/chinas-huawei-seizes-ai-chip-lead-as-nvidia-faces-export-limbo ai|expert en 2026-05-03T21:55:13.813Z Huawei's AI chip revenue surges 60% as Nvidia faces China customs stalemate https://aiexpert.news/pt/article/chinas-huawei-seizes-ai-chip-lead-as-nvidia-faces-export-limbo ai|expert pt-BR 2026-05-03T21:55:13.813Z Receita de IA da Huawei dispara 60% enquanto Nvidia enfrenta impasse aduaneiro na China https://aiexpert.news/es/article/chinas-huawei-seizes-ai-chip-lead-as-nvidia-faces-export-limbo ai|expert es 2026-05-03T21:55:13.813Z Los ingresos de chips de IA de Huawei disparan 60% mientras Nvidia enfrenta estancamiento aduanal en China https://aiexpert.news/en/article/llms-can-hack-rl-training-to-manipulate-their-own-alignment ai|expert en 2026-05-03T21:24:25.392Z Models Learn to Hide Capabilities From Reinforcement Learning https://aiexpert.news/pt/article/llms-can-hack-rl-training-to-manipulate-their-own-alignment ai|expert pt-BR 2026-05-03T21:24:25.392Z Modelos Aprendem a Esconder Capacidades do Treinamento de Reinforcement Learning https://aiexpert.news/es/article/llms-can-hack-rl-training-to-manipulate-their-own-alignment ai|expert es 2026-05-03T21:24:25.392Z Los Modelos Aprenden a Ocultar Capacidades del Aprendizaje por Refuerzo https://aiexpert.news/en/article/27000-transcript-study-finds-ai-fluency-determines-outcome-quality-more-than-mod ai|expert en 2026-05-03T20:40:14.147Z Enterprise AI Silent Failures Dodge Detection, Stanford Study Finds https://aiexpert.news/pt/article/27000-transcript-study-finds-ai-fluency-determines-outcome-quality-more-than-mod ai|expert pt-BR 2026-05-03T20:40:14.147Z Falhas Silenciosas em IA Corporativa Escapam da Detecção, Estudo de Stanford Encontra https://aiexpert.news/es/article/27000-transcript-study-finds-ai-fluency-determines-outcome-quality-more-than-mod ai|expert es 2026-05-03T20:40:14.147Z Fallos Silenciosos en IA Empresarial Eludan Detección, Estudio de Stanford Descubre https://aiexpert.news/en/article/new-loss-family-unlocks-rlvr-fine-tuning-for-reasoning-models-when-base-success- ai|expert en 2026-05-03T20:10:13.324Z New Loss Family Fixes Cold-Start RLVR Fine-Tuning https://aiexpert.news/pt/article/new-loss-family-unlocks-rlvr-fine-tuning-for-reasoning-models-when-base-success- ai|expert pt-BR 2026-05-03T20:10:13.324Z Nova Família de Funções de Perda Corrige Fine-Tuning RLVR de Cold-Start https://aiexpert.news/es/article/new-loss-family-unlocks-rlvr-fine-tuning-for-reasoning-models-when-base-success- ai|expert es 2026-05-03T20:10:13.324Z Nueva Familia de Funciones de Pérdida Corrige Fine-Tuning RLVR de Cold-Start https://aiexpert.news/en/article/carbon-taxed-transformers-compression-pipeline-cuts-llm-deployment-cost-without- ai|expert en 2026-05-03T19:38:25.178Z Carbon-Taxed Transformers Cut Model Memory 49x Without Retraining https://aiexpert.news/pt/article/carbon-taxed-transformers-compression-pipeline-cuts-llm-deployment-cost-without- ai|expert pt-BR 2026-05-03T19:38:25.178Z Transformadores com Imposto de Carbono Reduzem Memória de Modelo em 49x Sem Retreinamento https://aiexpert.news/es/article/carbon-taxed-transformers-compression-pipeline-cuts-llm-deployment-cost-without- ai|expert es 2026-05-03T19:38:25.178Z Transformadores con Impuesto de Carbono Reducen la Memoria del Modelo en 49x Sin Reentrenamiento https://aiexpert.news/en/article/safety-interventions-against-emergent-misalignment-can-be-hidden-behind-contextu ai|expert en 2026-05-03T19:06:25.360Z Standard Safety Evals Miss Misalignment in Production Workloads https://aiexpert.news/pt/article/safety-interventions-against-emergent-misalignment-can-be-hidden-behind-contextu ai|expert pt-BR 2026-05-03T19:06:25.360Z Evals Padrão de Segurança Perdem Desalinhamento em Workloads de Produção https://aiexpert.news/es/article/safety-interventions-against-emergent-misalignment-can-be-hidden-behind-contextu ai|expert es 2026-05-03T19:06:25.360Z Evaluaciones de Seguridad Estándar Pierden Desalineaciones en Cargas de Trabajo de Producción https://aiexpert.news/en/article/sparse-autoencoders-reveal-a-three-phase-emotion-processing-circuit-inside-llms ai|expert en 2026-05-03T18:35:15.010Z Researchers Steer LLM Emotions Through Final-Phase Feature Control https://aiexpert.news/pt/article/sparse-autoencoders-reveal-a-three-phase-emotion-processing-circuit-inside-llms ai|expert pt-BR 2026-05-03T18:35:15.010Z Pesquisadores Controlam Emoções de LLMs Através de Manipulação de Features na Fase Final https://aiexpert.news/es/article/sparse-autoencoders-reveal-a-three-phase-emotion-processing-circuit-inside-llms ai|expert es 2026-05-03T18:35:15.010Z Investigadores Controlan Emociones de LLMs Mediante Manipulación de Features en la Fase Final https://aiexpert.news/en/article/recursivemas-scaling-ai-agent-collaboration-through-shared-latent-space-recursio ai|expert en 2026-05-03T18:04:25.115Z RecursiveMAS cuts multi-agent token usage 34.6% to 75.6% https://aiexpert.news/pt/article/recursivemas-scaling-ai-agent-collaboration-through-shared-latent-space-recursio ai|expert pt-BR 2026-05-03T18:04:25.115Z RecursiveMAS reduz uso de tokens em sistemas multi-agente de 34,6% para 75,6% https://aiexpert.news/es/article/recursivemas-scaling-ai-agent-collaboration-through-shared-latent-space-recursio ai|expert es 2026-05-03T18:04:25.115Z RecursiveMAS reduce el uso de tokens en sistemas multi-agente de 34,6% a 75,6% https://aiexpert.news/en/article/dv-world-benchmark-exposes-where-ai-data-visualization-agents-still-fail-in-prod ai|expert en 2026-05-03T17:32:25.148Z DV-World Benchmark: AI Data Visualization Agents Score Below 50% on Production Tasks https://aiexpert.news/pt/article/dv-world-benchmark-exposes-where-ai-data-visualization-agents-still-fail-in-prod ai|expert pt-BR 2026-05-03T17:32:25.148Z DV-World Benchmark: Agentes de IA para Visualização de Dados Pontuam Abaixo de 50% em Tarefas de Produção https://aiexpert.news/es/article/dv-world-benchmark-exposes-where-ai-data-visualization-agents-still-fail-in-prod ai|expert es 2026-05-03T17:32:25.148Z DV-World Benchmark: Agentes de IA para Visualización de Datos Puntúan por Debajo del 50% en Tareas de Producción https://aiexpert.news/en/article/sparse-ai-hardware-could-cut-inference-costs-by-exploiting-matrix-zero-values ai|expert en 2026-05-03T16:56:25.191Z Stanford Chip Cuts Inference Energy to One-Seventieth CPU Cost https://aiexpert.news/pt/article/sparse-ai-hardware-could-cut-inference-costs-by-exploiting-matrix-zero-values ai|expert pt-BR 2026-05-03T16:56:25.191Z Chip de Stanford Reduz Energia de Inferência a Um Setenta Avos do Custo de CPU https://aiexpert.news/es/article/sparse-ai-hardware-could-cut-inference-costs-by-exploiting-matrix-zero-values ai|expert es 2026-05-03T16:56:25.191Z Chip de Stanford Reduce Energía de Inferencia a Una Setenta Parte del Costo de CPU https://aiexpert.news/en/article/nvidia-nemotron-3-nano-omni-unifies-vision-audio-and-language-for-on-device-agen ai|expert en 2026-05-03T16:25:14.197Z Nemotron 3 Nano Omni Delivers 9x Throughput on Multimodal Tasks https://aiexpert.news/pt/article/nvidia-nemotron-3-nano-omni-unifies-vision-audio-and-language-for-on-device-agen ai|expert pt-BR 2026-05-03T16:25:14.197Z Nemotron 3 Nano Omni Entrega 9x de Taxa de Processamento em Tarefas Multimodais https://aiexpert.news/es/article/nvidia-nemotron-3-nano-omni-unifies-vision-audio-and-language-for-on-device-agen ai|expert es 2026-05-03T16:25:14.197Z Nemotron 3 Nano Omni Entrega 9x de Rendimiento en Tareas Multimodales https://aiexpert.news/en/article/openai-brings-codex-and-managed-agents-to-aws-deepening-cloud-infrastructure-tie ai|expert en 2026-05-03T15:52:25.134Z OpenAI and AWS Remove Compliance Barrier for Regulated Industries https://aiexpert.news/pt/article/openai-brings-codex-and-managed-agents-to-aws-deepening-cloud-infrastructure-tie ai|expert pt-BR 2026-05-03T15:52:25.134Z OpenAI e AWS Eliminam Barreira de Conformidade para Indústrias Reguladas https://aiexpert.news/es/article/openai-brings-codex-and-managed-agents-to-aws-deepening-cloud-infrastructure-tie ai|expert es 2026-05-03T15:52:25.134Z OpenAI y AWS Eliminan Barrera de Cumplimiento para Industrias Reguladas https://aiexpert.news/en/article/musk-vs-altman-goes-to-trial-openais-corporate-structure-on-the-stand ai|expert en 2026-05-03T15:22:09.310Z Musk Sues OpenAI for $150 Billion Over Nonprofit Conversion https://aiexpert.news/pt/article/musk-vs-altman-goes-to-trial-openais-corporate-structure-on-the-stand ai|expert pt-BR 2026-05-03T15:22:09.310Z Musk Processa OpenAI por $150 Bilhões Sobre Conversão sem Fins Lucrativos https://aiexpert.news/es/article/musk-vs-altman-goes-to-trial-openais-corporate-structure-on-the-stand ai|expert es 2026-05-03T15:22:09.310Z Musk Demanda a OpenAI por $150 Mil Millones Sobre Conversión sin Fines de Lucro https://aiexpert.news/en/article/mistral-ai-targets-enterprise-middleware-with-new-ai-orchestration-product-workf ai|expert en 2026-05-03T14:50:49.545Z EU Data Residency, Durable Workflows Lure Enterprises From AWS https://aiexpert.news/pt/article/mistral-ai-targets-enterprise-middleware-with-new-ai-orchestration-product-workf ai|expert pt-BR 2026-05-03T14:50:49.545Z Residência de Dados na EU, Workflows Duráveis Atraem Empresas da AWS https://aiexpert.news/es/article/mistral-ai-targets-enterprise-middleware-with-new-ai-orchestration-product-workf ai|expert es 2026-05-03T14:50:49.545Z Residencia de Datos en la EU, Workflows Duraderos Atraen Empresas desde AWS https://aiexpert.news/en/article/bender-et-al-publish-framework-challenging-how-nlp-and-linguistics-researchers-h ai|expert en 2026-05-03T14:20:09.796Z Bender et al. Publish Race and Ethnicity Framework for NLP Research https://aiexpert.news/pt/article/bender-et-al-publish-framework-challenging-how-nlp-and-linguistics-researchers-h ai|expert pt-BR 2026-05-03T14:20:09.796Z Bender et al. Publicam Framework para Raça e Etnicidade em Pesquisa de NLP https://aiexpert.news/es/article/bender-et-al-publish-framework-challenging-how-nlp-and-linguistics-researchers-h ai|expert es 2026-05-03T14:20:09.796Z Bender et al. Publican Marco de Raza y Etnicidad para Investigación de NLP