Over the past few years, advanced machine learning algorithms have propelled artificial intelligence to remarkable feats in complex tasks and problem-solving. Yet, machines still falter at everyday activities. Neuroscientist Anthony Zador argues that modeling AI on animal brains could bridge this gap.
Artificial intelligence (AI) has much to gain from studying animal brains, according to Anthony Zador, a neuroscientist at Cold Spring Harbor Laboratory (CSHL). He believes neuroscience insights can help next-generation AI overcome key challenges.
Zador has dedicated his career to mapping brain neural networks from individual neurons. He began with artificial neural networks (ANNs), the foundation of today's AI boom, inspired by animal brain structures including ours.
In a Nature Communications article, Zador notes how enhanced algorithms enable superhuman AI performance in areas like chess and poker. Paradoxically, simple daily tasks remain elusive.
This gap could empower robots to master natural behaviors like hunting prey, nest-building, or even dishwashing—a "basic demand" Google CEO Eric Schmidt once deemed "extraordinarily difficult for a robot."
“Things we find difficult, like abstract thinking or chess, are easy for machines. Interacting with the physical world? That's hard. We've had half a billion years of evolution optimizing our neural circuits for effortless performance,” Zador explains.
Zador posits that rapid learning stems not from universal algorithms but from evolution-sculpted biological networks providing task-specific scaffolding, especially for survival essentials.
Related: Artificial Intelligence is Transforming Scientific Research Forever
"Squirrels leap between trees weeks after birth; mice don't. Squirrels are genetically wired for arboreal life," he notes.
This innate circuitry guides early learning. Less general than typical machine learning pursuits, adopting it in ANNs could yield future appliances that handle chores like clean dishes flawlessly.
