The rapid advancement of artificial intelligence is remarkable, but its energy cost presents a significant challenge. Training large models, constant data transfer, and operating data centers require immense electrical consumption. Current hardware architectures, which separate information storage and processing, are becoming inefficient for modern AI demands.

To address this, a team led by the University of Cambridge is proposing neuromorphic computing, inspired by the human brain. They have developed a new type of memristor that combines memory and processing in a way more akin to a synapse, promising drastically reduced energy consumption for AI hardware.

The developed memristor is based on a thin film of hafnium oxide doped with strontium and titanium, offering more stable and uniform behavior than previous designs. This stability is crucial for complex AI systems, as it allows for a more predictable, gradual, and repeatable response by relying on subtle changes at the material interface rather than unstable internal filaments.

Experiments suggest this approach could reduce the energy consumption of certain AI processes by up to 70%. While still in the laboratory phase with manufacturing challenges, this research signifies an important shift in AI hardware design, moving beyond simply increasing computational power to fundamentally redesigning how computation occurs, drawing inspiration from the brain's efficiency.

👉 Read the full article on Gizmodo en Español

(Content of the article was not accessible. This summary is based on the title and general knowledge.)

The number of AI health tools has dramatically increased in recent years, with applications spanning diagnostics, treatment planning, and personalized medicine. These tools aim to analyze vast amounts of medical data and identify patterns, offering insights that human experts might miss.

However, significant questions remain regarding their efficacy and reliability. While many AI tools demonstrate high performance on specific datasets, their ability to maintain this performance across diverse patient populations and complex real-world clinical scenarios requires thorough validation. Concerns about bias and fairness across different demographic groups are also prominent.

The article likely discusses how, amidst the proliferation of AI health tools, their clinical utility, safety, and the evolving regulatory frameworks need to be critically examined. Rigorous validation and transparent evaluation are crucial for these technologies to gain widespread trust and adoption in healthcare settings.

👉 Read the full article on Technology Review

While much attention is focused on visible digital power elements like satellites, artificial intelligence, and chips, the United States is engaged in a quieter, yet crucial, strategic game involving the submarine cables that underpin the internet. These cables form the physical backbone of the global digital communication network, carrying the vast majority of online traffic.

The article highlights the Humboldt cable project, driven by Chile and Google, as a prime example. This project, connecting Chile directly with Oceania via the South Pacific, is more than just an improvement in connectivity; it's a significant piece in the geopolitical competition between the United States and China. It aims to streamline data traffic between South America and Asia-Pacific, which previously relied on less efficient routes, and establish Chile as a regional digital hub.

The Humboldt cable is positioned as an alternative to a similar trans-Pacific route proposed by China. Washington is wary of China's expanding technological influence in Latin America, and strategic infrastructure like submarine cables is particularly sensitive due as they are critical arteries for data, commerce, and governmental communications.

This project underscores that the battle for digital power is not only fought in laboratories or data centers but also silently beneath thousands of kilometers of saltwater. The laying of new cables not only enhances speed and capacity but also redistributes influence, dependence, and digital sovereignty on a global scale.

👉 Read the full article on Gizmodo en Español