yoshua bengio

shifts from deep learning to neurosymbolic hybrids, seeking architectures that combine system 1/2 reasoning has shifted focus from pure deep learning to neurosymbolic architectures that combine neural pattern recognition with symbolic reasoning systems Bengio hypothesizes that human-like AI requires disentangled representations - sparse, high-level abstractions mirroring how consciousness compresses sensory data into interpretable concepts. This challenges pure end-to-end deep learning by emphasizing the need for structured world models.