Pathway’s brain-inspired architecture development on Amazon SageMaker HyperPod

Pathway has developed BDH, a brain-inspired AI architecture that performs reasoning in latent space without generating intermediate text, overcoming limitations of transformer models. BDH updates internal memory during inference, enabling iterative problem-solving without retraining. The model integrates with PyTorch and scales training via Amazon SageMaker HyperPod.