Modelos, agentes, investigación y negocio de la IA.
A Databricks analysis shows that measuring AI cost only by token is misleading: the price per task reveals that seemingly expensive models can be more cost-effective. Additionally, the intermediary software (harness) impacts the final cost by up to 3 times.
Envision CEO Lei Zhang unveiled the 'Mission Gobi' plan at VivaTech, aiming to develop 5 GW of green AI capacity in deserts by 2030. The initiative reverses traditional logic: instead of bringing energy to data centers, it brings data centers to where abundant renewable energy is, thus avoiding competition with homes and industries.
Apple has released the first public betas of watchOS 27 and iOS 27, with Siri AI as the main highlight. Users can test features like a smarter assistant, one-handed gestures, and a proactive Smart Stack. Here's what's new and why it matters.
SK Hynix, the largest supplier of high-bandwidth memory (HBM) for Nvidia, debuted on Wall Street on Friday with a record $26.5 billion raise, surpassing Alibaba as the largest IPO by a foreign company in the US. The company, which briefly overtook Samsung as South Korea's most valuable firm, capitalizes on the AI boom.
Although 61% of organizations already use AI agents, almost none deploy them in production due to lack of trust. The real obstacle is not technology, but the absence of governance controls such as isolation, least privilege, and real-time monitoring.
Microsoft reported a 25% increase in its net carbon emissions during 2025, reaching 20 million tons of CO2 equivalent. The main cause is the massive construction of data centers to support its artificial intelligence services. The company says it maintains its 2030 target, but analysts doubt its viability.
Despite the rise of AI agents, fundamental capabilities such as identity, reliable execution, and intention remain unsolved problems. This article examines why these challenges are critical for enterprise adoption and how some platforms are beginning to address them.
OpenAI has introduced ChatGPT Work, an autonomous agent integrated into ChatGPT capable of executing actions across multiple applications and files for hours. This launch marks a shift from conversational response to task execution, with profound implications for workplace productivity and process automation.
OpenAI has announced the closure of ChatGPT Atlas, its AI-powered web browser, just nine months after its launch. The decision, effective August 9, ends an experiment that aimed to turn ChatGPT into a superapp but failed to gain traction against established competitors like Chrome and Edge.
Researchers from MIT and the nonprofit organization Thorn have developed an auditing technique that identifies whether a generative model has been fine-tuned to create child sexual abuse material (CSAM) without having to generate it. The method analyzes the model's internal representations and achieved 100% accuracy in tests. This could transform moderation on open-source model platforms and close a critical security gap.
Economists at Northwestern University recalculated exposure to AI considering the entire workforce, not just technology users. The results show that secretaries, assistants, and payroll employees top the risk list, far ahead of programmers.
The Apple Car never went into production, but its development spurred the creation of the Neural Engine, the heart of artificial intelligence in Apple devices. This article analyzes how a failed project transformed the company's chip strategy and its impact on the AI market.