AI Learns Basic Skills: Kindergarten Approach
- artificial intelligence, like children, may benefit from a "kindergarten curriculum" approach too learning.
- The findings, published in Nature Machine Intelligence, detail how recurrent neural networks (RNNs) demonstrate enhanced performance when trained using this staged learning method.Researchers at New York University (NYU)...
- Cristina Savin, an associate professor at NYU's Center for Neural Science and Center for Data Science, explained the concept.
AI evolves: A “kindergarten curriculum” for AI dramatically boosts its grasp of complex problems. Researchers at New York University (NYU) discovered that teaching AI simple cognitive tasks first significantly enhances its ability to learn and perform advanced functions. This innovative “kindergarten approach” uses staged learning, mirroring how humans build upon basic skills. The team, publishing in Nature Machine Intelligence, found that Recurrent Neural Networks (RNNs) showed a marked improvement in learning speed when following this method. This breakthrough, detailed by lead researchers like Cristina Savin, has promising implications for machine learning and artificial intelligence. RNNs excel in speech recognition and language translation; refining their training could lead to exciting technological advancements—something News Directory 3 will be watching. Discover what’s next in the world of AI’s rapid advancement.
AI Learns Faster with Kindergarten-Style Training Approach
artificial intelligence, like children, may benefit from a “kindergarten curriculum” approach too learning. new research indicates that training AI on simple cognitive tasks initially can significantly improve its ability to tackle more complex challenges later on. this method enhances AI training and machine learning capabilities.
The findings, published in Nature Machine Intelligence, detail how recurrent neural networks (RNNs) demonstrate enhanced performance when trained using this staged learning method.Researchers at New York University (NYU) spearheaded the study.
Cristina Savin, an associate professor at NYU’s Center for Neural Science and Center for Data Science, explained the concept. “From very early on in life, we develop a set of basic skills… With experience, these basic skills can be combined to support complex behavior,” Savin said.”Our work adopts these same principles in enhancing the capabilities of RNNs.”
RNNs are particularly useful in speech recognition and language translation. However, existing training methods often struggle with complex cognitive tasks, failing to replicate aspects of human and animal behavior. The NYU team, including David Hocker and Christine Constantinople, sought to address this limitation.
The team drew inspiration from experiments involving laboratory rats. The rats learned to associate sounds and light cues with the availability of water in a compartmentalized box. the animals needed to combine these simple associations to successfully retrieve the water.
The researchers then applied these principles to train RNNs, tasking them with a wagering game that required building upon basic decision-making skills. The “kindergarten curriculum learning” approach was then compared to existing RNN training methods. The AI training results showed the new method improved learning speed.
“AI agents first need to go through kindergarten to later be able to better learn complex tasks,” Savin observed.
The National Institute of Mental Health, the State of New York, the Simons Foundation, and the Secunda Family Foundation supported the research.
what’s next
The researchers suggest that a more holistic understanding of how past experiences influence the learning of new skills could further improve AI systems.