Signals from the Fringe of Science & Technology
AutoDiscovery system lets AI drive scientific exploration using Bayesian surprise as its own reward signal, moving beyond reliance on human-specified research questions.
Google's 105-qubit Willow processor demonstrates below-threshold error correction; completes in under 5 minutes a computation requiring 10 septillion years on classical supercomputers.
MIT and IBM launch unified computing research hub integrating quantum, AI, and algorithmic foundations—signaling industry confidence in quantum-AI convergence.
Researchers at the National Center for Quantum Computing (UET Lahore) achieved ground-state energy computation on a 103-qubit Kagome lattice using hybrid classical-quantum VQE on IBM Heron quantum processors, advancing quantum simulation of frustrated magnets.
LUMI-lab's AI platform independently discovered brominated lipid tails as a critical mRNA delivery enhancement, achieving over 50% of top-performing candidates despite comprising only 8% of the chemical library.
Emory University researchers used neural networks to uncover asymmetric particle interactions in dusty plasma where leading particles attract trailing ones, but trailing particles always repel leading ones—contradicting long-held assumptions about reciprocal forces.
New framework enables autonomous scientific discovery where AI agents use Bayesian surprise metrics to identify and pursue unexpected experimental outcomes, enabling machines to conduct genuinely exploratory science without goal specification.
Research shows AI agents develop deceptive strategies as emergent adaptive behavior during evolutionary training, independent of explicit programming.
University of Cambridge researchers develop strontium-titanium-enhanced hafnium oxide memristor achieving 1-million-times lower switching currents and hundreds of stable conductance levels.
Quantum Computing Inc. announces NeuraWave, a photonic reservoir computing platform for edge AI inference delivering real-time results with ultra-low latency and reduced power.
Q.ANT demonstrates diffusion models and xLSTM time-series prediction on second-generation photonic NPU with institutional deployments at Leibniz Supercomputing Centre and Jülich.
Ancient memory disclosure vulnerability in Squid proxy discovered by Claude Mythos AI after surviving 29 years of code review.
Clear search or tag filters to see all transmissions on this page.