Artificial intelligence systems are modeled after the human brain, but a new field of research at Columbia University in New York is looking at how the living brain works and how it can be improved in the development of AI. I’m researching if it might contain any clues.
Columbia University is one of seven universities designated by the National Science Foundation as new national AI research institutions, and the $20 million received by Columbia University will strengthen the university’s Artificial and Natural Intelligence AI Institute (ARNI) will be The goal is to conduct research that “ties major advances in AI systems to revolutionize understanding of the brain.”
Columbia University computer science professor Richard Zemmel told Fox News Digital that the goal is to bring together top researchers in AI and neuroscience to cross-train AI systems and people to benefit.
“The idea is to go both ways,” Zemmel said. “AI is inspired by the brain, and neural nets have something loosely connected to the brain.”
AI-Powered ‘Life-saving Radio’ Helps Surgeons Improve Efficiency and Accuracy
A digital image of the brain in the palm of your hand using artificial intelligence technology. (St. Petersburg)
One of the central ideas behind AI was to mimic the structure of the brain in hopes of creating something akin to a thinking machine. Modeled after the brain, artificial neural networks consist of millions of processing nodes that help AI systems learn when presented with data.
The “transformer” neural network, which has been in use for the past five years or so, aims to get even closer to the human brain by focusing on the context of the question asked to arrive at a more accurate answer. Zemmel said Transformers focuses on the concept of “attention.”
“It’s called the cocktail party effect,” says Zemmel. “You’re at a party and you’re barely deaf, but you can hear your name even though there’s a lot of conversation going on. But somehow your brain recognizes something. You can pay attention to it.”
He said this notion of “attention” makes generative AI output increasingly accessible to people asking questions about AI systems. This kind of research has opened the door to the question of whether improvements in AI can help researchers better understand the brain.
“Does understanding these complex neural networks lead to hypotheses and new things to explore in the brain?” said Zemmel.
AI tools help doctors make sense of chaotic patient data and identify diseases: ‘more meaningful’ interactions
Professor Richard Zemel teaches computer science at Columbia University in New York, NY. (Columbia University)
A major issue that Columbia University considers involves understanding the concept of “robust and flexible learning.” He said that so far many AI systems do well at certain tasks but perform poorly when given other tasks, while the human brain exhibits greater adaptability.
But AI has shown that it can develop language skills quickly, which is an example of AI talent that could help us understand how to train the human brain more efficiently, Zemel says. said Mr.
“Many of these new systems recognize new language tasks very well. With just one or two examples, the system learns something very quickly. I will learn,” he said. The question is, “So do we get an idea of what we might want to do in a different way when it comes to training humans?”
Another area is continuous learning. It addresses the question of when and how both humans and AI systems forget information, and how that information can be recalled.
“AIs can sometimes suffer from massive amounts of forgetfulness,” says Zemmel. “Both are troubled in many ways, so this is a good area to study and great for figuring out if there’s a way they can both help each other.”
A third crossover problem that affects both humans and AI systems is the uncertainty principle.
AI Healthcare Platform Predicts Diabetes Accurately, But ‘No Substitute for Patient Care’
Columbia University Library building at night, Manhattan, New York, USA. (Photo credit: Peter Titmuss/Education Images/Universal Images Group via Getty Images)
“Many AI systems out there are not very good at recognizing uncertain times when they should be uncertain,” he said. “And people aren’t very good at it.”
Practical applications of this kind of cross-training between AI and human brains have already been developed and refined. One example is a kind of “brain-machine interface” that helps build smarter prosthetic devices for people, such as mechanical arms for those who can’t control their arms.
CLICK HERE TO GET THE FOX NEWS APP
Zemmel said “AI-assisted prosthetic devices” are being developed that allow movement partly through the brain and partly through an AI interface.
He said it is hoped that AI and neuroscience experts and Columbia will continue to make these kinds of connections.
“We’re bringing these people together, in the same room, coming up with ideas, going back and forth, trying to figure out what to test and explore,” Zemmel said.
