Sam Altman told the United Nations Security Council that AI policy cannot be made by labs in San Francisco alone. Here is what he asked governments to build, which claims rest on evidence and which rest on prediction, and what the speech leaves unresolved for democratic institutions.
OpenAI expanded OpenAI Academy with role-based learning paths and course badges, and launched a Grab partnership to train 30,000 Southeast Asia gig and merchant partners. The training is free and substantive, but vendor-run curricula also define what competent and safe AI use means. This analysis examines what the courses teach, what they leave out, and why no outcomes evidence exists yet.
A eukaryotic amoeba found in Lassen Volcanic National Park reproduces at 145 degrees Fahrenheit, breaking the previous 140-degree ceiling for complex life and widening the conditions under which NASA will search for life beyond Earth.
An ETH Zurich experiment rapidly reduced uncertainty in one component of a levitated particle’s angular motion. Technical noise and a short cavity-limited lifetime still constrain its use.
Heat-driven modules passed an inflation wave around a wired ring and moved a laboratory crawler 25 millimeters, while slow cycles and thermal failures bound the result.
Remote entanglement carried error syndromes between two ion-trap modules, enabling detection and active correction. The small experiment remains well below fault-tolerant performance.
A four-photon experiment identified logical Bell states with 70.8 percent average success. The gain could help photonic networks and computers, but it was postselected and did not provide full error correction.
An ETH Zurich controller learned how finger commands moved a grasped pen without simulation training or demonstrations, achieving a reported 0.64-millimeter mean in-plane error in laboratory tests.
A lithium-niobate processor combines spatial and time-bin mixing at 4 gigahertz. Its scale is substantial, but its quantum-advantage and AI results remain author-reported benchmarks.
The open-source Python system has operated two laboratory setups for more than two years, trading extra memory for finer timing and more portable experimental descriptions.
A Warsaw team trapped more than a million atoms of each species and mapped 14 resonances in a related isotope mixture. Rapid potassium-41 losses still block the spectroscopy needed for molecule assembly.
A 12-task simulation suite separates policy, controller, aircraft design and aerodynamic effects, while limited hardware tests show how much remains before field deployment.
Shallow compiled circuits exposed three signatures of topological order, but their 97.9 to 99.0 percent overlaps were classical calculations, not hardware-state fidelities.
A first-party MIT and OpenAI case study reports researcher intervention to improve four of 40 target measurements on four fixed-frequency qubits. Noisy tunable-qubit scans still required expert judgment.
A controlled study found that offline predictions ranked state estimators more like feedback performance when they reproduced the robot’s measurement-update schedule.
A jamming palm and adaptive fingers improved screwdriver force resistance in laboratory tests, while uneven tool results show the distance to industrial deployment.