A new AI model demonstrates significant improvements in processing complex, multi-step reasoning tasks by integrating a ‘chain-of-thought’ approach directly into its training process. Unlike previous models that required explicit prompting for step-by-step solutions, this system is trained to automatically decompose problems, showing enhanced performance on benchmarks in mathematics, coding, and logical deduction. Researchers highlight that this method reduces errors and increases transparency in the AI’s decision-making path. The development suggests a shift towards more inherently structured and reliable reasoning in next-generation AI systems. Read the full article at https://technologyreview.com/2023/10/ai-reasoning-breakthrough.