NVIDIA Unveils RLP: A New Era of Reinforcement Learning for AI Pretraining

NVIDIA researchers have developed a groundbreaking training objective called Reinforcement Learning Pretraining (RLP), which integrates reinforcement learning directly into the pretraining stage of AI model development. The core concept of RLP treats a short chain-of-thought as an action that is rewarded based on the information gain it provides for the model’s next-token prediction. This innovative method creates a verifier-free and dense reward system that can be applied at scale to standard text streams. Unlike previous methods that depend on sparse signals or external verifiers, RLP’s approach allows for updates at every single token position. Experiments demonstrate that RLP can yield significant improvements in a model’s reasoning capabilities, with one model showing a substantial increase in its average score on scientific reasoning benchmarks.

Sources:

How AI and Machine Learning are Revolutionizing Railroad Safety by Predicting Trespasser Hotspots

New applications of artificial intelligence and machine learning are being used to enhance railroad safety by identifying and predicting trespassing hotspots. By analyzing anonymized location data from devices within geofenced railway corridors, these advanced systems can detect areas where individuals frequently ’linger’—a key indicator of trespassing. This proactive approach allows railroad operators to intervene before incidents occur, aiming to reduce injuries and fatalities. The implementation of these AI and ML solutions enables a more strategic allocation of safety resources to the areas of greatest risk. Future developments in this technology are focused on moving from historical analysis to the active, real-time detection of trespassing events.

Tensor Unveils the Robocar: The World’s First Fully Autonomous Personal Robot Car

At GITEX Global 2025, US-based Tensor Auto has unveiled the Tensor Robocar, which it bills as the world’s first fully autonomous personal robot car. The company states the vehicle is a level-four autonomous car, meaning it does not require a driver’s license or special insurance for operation. The Robocar was designed from the ground up for full autonomy, not as a traditional vehicle retrofitted with smart systems. It is equipped with over 100 sensors and cameras, including five Lidar sensors, to perceive its surroundings with high precision, enabling it to navigate and park without any human intervention. The vehicle’s processing power is reportedly up to 8,000 computing units, a significant leap compared to conventional smart cars.

Sources:

Oracle Launches Enterprise AI Data Platform to Connect Generative AI with Business Data

Oracle has announced the general availability of the Oracle AI Data Platform. This platform is engineered to help customers securely connect generative AI models with their proprietary enterprise data, applications, and workflows. By combining automated data ingestion, semantic enrichment, and vector indexing with built-in generative AI tools, the platform streamlines the entire process from raw data to production-grade AI. The Oracle AI Data Platform leverages the combined power of Oracle Cloud Infrastructure (OCI), Oracle Autonomous AI Database, and the OCI Generative AI service. It also integrates NVIDIA accelerated computing infrastructure to manage high-performance workloads, aiming to accelerate innovation and productivity across all business lines.

Sources:

Oracle Boosts AI Capabilities with New OCI Dedicated Region and AMD GPU Supercluster Partnership

Oracle has announced OCI Dedicated Region25, a new, smaller-footprint option for its cloud infrastructure that can be deployed in customer data centers in as few as three racks. This offering provides access to over 200 AI and cloud services while helping organizations meet strict data sovereignty requirements. In a separate major announcement at Oracle AI World, Oracle revealed a partnership with AMD to build an AI supercluster. Oracle Cloud Infrastructure will deploy 50,000 AMD Instinct MI450 Series GPUs, starting in the third quarter of 2026. This strategic move positions Oracle as the first hyperscaler to offer a publicly available AI supercluster of this scale with AMD’s latest AI accelerators. The new OCI superclusters will be based on the AMD Helios rack design, featuring next-generation EPYC CPUs and Pensando networking.

Sources:

D-Wave Becomes Founding Member of Italy’s ‘Q-Alliance’ to Build Quantum Computing Hub

D-Wave Quantum Inc. has been announced as a founding member of the Q-Alliance, an initiative to establish a world-leading quantum computing hub in Lombardy, Italy. The alliance, formalized via a Memorandum of Understanding, aims to accelerate scientific discovery, industrial transformation, and Italy’s digital sovereignty by leveraging D-Wave’s quantum computing technology. Supported by the Italian government’s strategic framework for digital and quantum technologies, the Q-Alliance will function as an open and inclusive ecosystem for Italy’s scientific community, academia, and industry. The initiative will also focus on nurturing young talent through scholarships and training programs to build a quantum-ready workforce.

Sources:

IBM and Basque Government Launch Europe’s First Quantum System Two in Spain

The Basque Government and IBM have inaugurated Europe’s first IBM Quantum System Two in Donostia-San Sebastián, Spain. The system is powered by a 156-qubit IBM Quantum Heron processor, one of the company’s highest-performing quantum processors to date. This installation marks a significant milestone for the Basque Country’s strategy to become a leader in the emerging quantum technology field. The new quantum computer is designed for scalability and to run utility-scale algorithms that are beyond the capacity of classical supercomputers. It will be accessible to members of the IBM-Euskadi Quantum Computational Center to advance research in key areas such as energy, biomedicine, materials science, and artificial intelligence.

Sources: