Artificial intelligence has made tremendous progress by mastering language, diagnosing diseases and solving protein structures, yet we still don’t have robots that can wash our dishes. Why? Neuroscientist Anthony Zador, MD, PhD, argues that the answers lie in understanding the brain itself.
Zador, the Alle Davis Harris Professor of Biology at Cold Spring Harbor Laboratory and a leading expert at the intersection of neuroscience and artificial intelligence, delivered the 27th annual Swartz Foundation Mind Brain Lecture at Stony Brook University’s Staller Center on March 31. Zador was named one of Foreign Policy’s 100 Leading Global Thinkers and is a recipient of the Brain Research Foundation Fellowship, the Gill Symposium Transformative Investigator Award and the Allen Distinguished Investigator Award.
His talk, A Neuroscientist’s Guide to Artificial Intelligence, explored how unraveling the brain’s computations can fuel AI breakthroughs and how AI can help us decode the mysteries of the mind.
“My lab actually does both neurobiology research — with actual organisms — and computational work in AI,” Zador explained. “But today, I’m going to focus on the AI side.”
The event, hosted by Stony Brook’s Department of Neurobiology and Behavior and supported by the Swartz Foundation, continued the tradition of bringing cutting-edge discussions on the brain to a broad audience.
For decades, science fiction promised us intelligent machines, from Star Trek’s Data to HAL 9000 in 2001: A Space Odyssey. Many of these imagined AI systems were expected to be just around the corner. But as Zador pointed out, reality has lagged behind.
“When I was a kid, I watched a lot of sci-fi,” he said. “And it was just assumed that in the not-too-distant future, we’d have intelligent AI controlling robots. Well, we’re now well beyond some of those timelines — and we still aren’t there.”
Despite major AI advances, today’s systems still fail in surprising ways. While large language models like ChatGPT can mimic human conversation convincingly enough to pass the Turing Test (a classic measure of machine intelligence), they also generate strange mistakes. AI vision systems, he pointed out, can recognize cows in grassy fields but may misidentify them as camels when placed on a beach.
“These kinds of errors reveal something fundamental,” Zador explained. “AI doesn’t generalize the way humans do. It gives more weight to patterns in the training data than to deeper, more flexible reasoning.”
The problem, according to Zador, stems from an assumption about intelligence itself. Many believe that tasks like playing chess or solving math problems represent the pinnacle of cognitive ability. But in reality, the most complex tasks — like walking through a crowded street or recognizing a friend’s face in different lighting — are things that even infants can do easily.
This idea is explained by Moravec’s Paradox, a principle in AI research that suggests that the hardest tasks for machines are often the ones that come naturally to humans and animals.
“We are all prodigious Olympians in perceptual and motor areas,” roboticist Hans Moravec once wrote, “so good that we make the difficult look easy. Abstract thought, though, is a new trick, perhaps less than 100,000 years old.”
Zador emphasized that our brains are shaped by millions of years of evolution. This means that the fundamental abilities shared by all animals — perception, movement and intuition — are actually the result of the brain’s most intricate computations. This explains why AI can defeat a world champion at chess but still struggles to open a refrigerator or fold laundry.
AI companies have aimed to address these shortcomings by scaling up and training models on ever-larger datasets and refining algorithms. But Zador argued that this approach has limits.
“Just because airplanes improved over the past century doesn’t mean they’ll ever fly to the moon,” he said. “There are fundamental constraints to how far brute-force scaling can take us.”
Instead, he suggested that neuroscience may hold the key. The brain doesn’t learn by absorbing trillions of data points; it develops through experience, forming structured representations of the world. If AI could be designed to learn in the way humans and animals do, it might overcome some of its most frustrating blind spots.
Zador is one of the pioneers of NeuroAI, a growing field that seeks to merge insights from neuroscience with artificial intelligence research. By reverse-engineering the brain, scientists hope to build machines that learn more efficiently and reason more like humans.
This cross-disciplinary approach has already yielded breakthroughs. Neuroscience-inspired architectures, such as transformer models, have propelled AI forward in natural language processing. Meanwhile, AI tools are helping neuroscientists map the brain’s circuits in unprecedented detail.
The exchange of ideas between these fields, Zador suggested, will only deepen in the years ahead. AI and neuroscience are on the cusp of transforming one another and reshaping our understanding of intelligence itself.
“Ultimately, the goal is not just to build smarter AI,” he said. “It’s to unlock the secrets of the mind.”
— Beth Squire
Add comment