Modelos, agentes, investigación y negocio de la IA.
Hugging Face has released a major update to its kernels, optimizing key operations like attention and normalization. This translates to faster training, lower memory consumption, and better support for recent hardware, solidifying its platform as a reference in AI.
The Salamanca City Central School District in New York has launched a pilot program using the Realbotix M-Series humanoid robot and the Optio tutoring platform. The robot, equipped with conversational AI and facial expressions, will act as an assistant in robotics and AI classes, while Optio provides 24/7 personalized tutoring. The project starts with high school students and could expand to 500 students by fall 2026.
The collaboration between Microsoft and the FBI led to the arrest of Peter Stokes, an alleged member of Scattered Spider, thanks to the Global Device Identifier (GDID) of Windows 11. The case reveals how Microsoft's telemetry becomes a criminal investigation tool, but also raises privacy questions.
Hon Hai (Foxconn) reported a 40% increase in quarterly sales, beating expectations, thanks to strong demand for AI servers. AI rack shipments will continue their momentum this quarter, solidifying the company as a key player in Nvidia's supply chain.
Singapore authorities seized a $42 million mansion and froze bank accounts of suspects accused of smuggling Nvidia GPUs to China, evading U.S. export restrictions. The case, linked to the DeepSeek investigation, reveals how Singapore was used as a transit point to bypass sanctions.
Jane Fraser, CEO of Citigroup, stated that banking is fighting two simultaneous AI races: an offensive one to boost revenue and a defensive one to manage risks. This vision redefines the global financial sector's strategy.
Nvidia has launched a business model where, in addition to selling hardware, it receives a percentage of revenue generated by its AI cloud customers. In return, it offers credit tokens and financial support to startups with limited capital. The first partners are Sharon AI and Firmus Technologies.
A UNICEF analysis across 10 countries estimates that 20 million children already use artificial intelligence tools, adopting them three times faster than adults. This phenomenon, described as a 'global experiment,' outpaces the response capacity of current regulatory frameworks.
According to the Financial Times, OpenAI is pushing a proposal in which leading artificial intelligence companies would cede a 5% equity stake to the U.S. government. The move aims to share the benefits and risks of AI at the national level, but raises doubts about its feasibility and potential conflicts of interest.
Meta has announced that Ray-Ban smart glasses will require a monthly subscription to access advanced features like real-time translation and an enhanced AI assistant. This move reflects a growing industry trend toward recurring revenue models beyond hardware sales.
WhatsApp has started rolling out usernames, a feature that allows identification without sharing your phone number. However, early tests reveal that names like 'indiamodi' or 'cz_binance' are available or have already been reserved, triggering alarms about identity theft. The Indian government has requested a halt to the rollout until risks are assessed.
Alibaba has banned its employees from using Claude Code, Anthropic's AI coding tool, deeming it high-risk. The decision reflects growing geopolitical tensions and strict Chinese AI regulations, with implications for the global adoption of code assistants.