The AI ​​brain that’s getting smarter by shrinking – Neuroscience News

Summary: In the world of AI, bigger is usually considered better – but this leads to enormous energy consumption and computing costs. Taking a cue from human biology, a research team has developed a brain-inspired “selective pruning” framework for spiking neural networks (SNNs).

The study shows that AI no longer needs more connections to learn complex tasks; it takes the right ones. By mimicking the way an infant’s brain strengthens long-distance connections while “eliminating” local clutter, this new AI achieves continuous learning—mastery of perception, motor control, and interaction—while actually becoming smaller and more energy efficient over time.

Important facts

• The “Toddler” Approach: The human brain doesn’t just add connections; they refine them. This model follows a “simple to complex” progression, where primary modules (such as perception) are matured before moving on to higher cognitions.
• Selective cutting: Unlike traditional AI, which freezes weights to prevent forgetting, this system…