How Close Are Today’s AI Models to AGI—And to Self-Improving into Superintelligence?

General-purpose humanoid robots and embodied AI continue to face significant challenges in real-world applications, primarily due to their inability to replicate the physical intuition that humans develop through experience. This limitation has resulted in a slow progression in the deployment of these technologies despite advancements in hardware.

New photographs have captured moths engaging in the unusual behavior of sipping tears from a moose, marking only the second observation of this phenomenon outside the tropics. This behavior, previously noted primarily in tropical regions, highlights a unique interaction between species in different ecosystems.

A recent study has introduced a verification framework aimed at detecting model weight exfiltration from AI inference servers, particularly focusing on the risks posed by attackers who may hide sensitive information within model outputs. This framework formalizes model exfiltration as a security game and evaluates its effectiveness on various open-weight models, including the MOE-Qwen-30B.

A new automated pipeline has been developed to enhance the diagnosis of coronary artery disease (CAD) by predicting blood pressure distributions using coronary computed tomography angiography (CCTA). This system utilizes computational fluid dynamics (CFD) simulations to generate consistent training data while reducing the manual workload associated with traditional methods.