As the energy consumption of neural networks continues to grow, different approaches to deep learning are needed. A neuromorphic method offering nonlinear computation based on linear wave scattering can be implemented using integrated photonics.

I watched United Nations delegates debate AI-based weapons that can fire without human initiation. Humans cannot be taken out of that decision-making.

By Kate Graham-Shaw

Imagine a weapon with no human deciding when to launch or pull its trigger. Imagine a weapon programmed by humans to recognize human targets, but then left to scan its internal data bank to decide whether a set of physical characteristics meant a person was friend or foe. When humans make mistakes, and fire weapons at the wrong targets, the outcry can be deafening, and the punishment can be severe. But how would we react, and who would we hold responsible if a computer programmed to control weapons made that fateful decision to fire, and it was wrong?

The authors identify reusable ‘dynamical motifs’ in artificial neural networks. These motifs enable flexible recombination of previously learned capabilities, promoting modular, compositional computation and rapid transfer learning. This discovery sheds light on the fundamental building blocks of intelligent behavior.

Researchers at the University of Tokyo developed a method to integrate engineered skin tissue with humanoid robots, enhancing mobility, self-healing, sensors, and realism.