Quantum computing advances through IBM research and AI model compression

Researchers from IBM and the University of Chicago have achieved a milestone in quantum computing by completing a calculation in approximately 15 minutes that would be practically impossible for the most advanced classical computers to replicate. The experiment utilized a technique related to random circuit sampling, which is typically used to test the limits of quantum hardware. To address the difficulty of verifying results in such complex calculations, the team developed a method that introduces a structure allowing for error detection during the process.

The system employed 70 logical qubits designed to protect quantum information from noise and operational errors. This approach allowed the researchers to perform calculations beyond the reach of classical systems while providing evidence that the results are reliable.

In a separate development within the quantum sector, the Spanish deep tech company Multiverse Computing is utilizing quantum-inspired techniques to address the high costs of executing artificial intelligence models. The company has developed software that can compress large language models (LLMs) by at least half without sacrificing precision. This compression reduces memory requirements, allowing data centers to double their workload without increasing hardware, energy consumption, or cooling needs. This technology has drawn interest from major semiconductor firms such as Intel and Qualcomm.