AI needs 92,000 years to learn language as well as... a child

A new study estimates that it would take an AI system like ChatGPT 92,000 years to reach the same language-learning speed as a normal child.

The fact that children learn language more effectively than any current artificial intelligence (AI) system could shape the future of education , technology and brain science - Illustration photo

New research from Professor Caroline Rowland at the Max Planck Institute for the Psychology of Languages ​​(Netherlands) shows that children learn language much more effectively than current AI technology.

Children not only absorb language but also build their own systems through interactions, emotions and vivid experiences. Meanwhile, AI technology still has difficulty connecting multi-sensory information.

This research not only helps understand language development in children but also opens up new directions for improving AI technology in the future.

Children "live in language", AI just "processes data"

According to Professor Rowland, children "live in language", while AI only "processes data". It is the active participation in the world : from crawling, touching, listening, seeing, to asking and imitating that helps the child's brain naturally connect language with emotions, gestures and context.

In fact, research estimates that it would take an AI system like ChatGPT 92,000 years to reach the same language learning speed as a normal child.

Scientists have found an explanation for why children learn languages ​​much faster and more effectively than artificial intelligence (AI) systems. Accordingly, the human brain possesses special learning mechanisms that machines have not yet been able to simulate, focusing on three key differences: the way information is received, social interaction and the mechanism of language construction.

Children learn language not only from text data, but also combine information from multiple senses at once: hearing, sight, touch, even smell and taste. For example, when a baby learns the word “dog,” the brain simultaneously remembers the sound of a dog’s bark, the image of a dog, the feel of its soft fur, and the happy feeling of playing with it.

It is this multi-layered combination that helps children understand and remember language deeply, creating connections between sounds, images, emotions and meanings. Meanwhile, current AI systems still mainly process static text data and lack the ability to connect information from multiple senses, which makes it difficult for machines to achieve natural understanding like humans.

Different "learning" contexts

AI technology, despite its intelligence, is still far behind humans in learning languages ​​- Photo: AI

Another key factor is that children learn language in vivid contexts. When parents read books, point to a bird in the sky, or play with friends, children are constantly taking in information from sounds, images, gestures, and emotions. This helps the brain form rich networks of connections, thereby learning and using language more naturally.

In contrast, AI learns from static text data and lacks the emotional, gestural, and social nuances that are core components of human language expression and understanding.

Besides, children do not "pre-load" knowledge like AI but build their own language system through trial and error.

For example, children can add words expressing an action in the past tense without being formally taught, such as the word "đà…rổi": "Con đã ăn cơm" (I have eaten rice already). This is a step-by-step process, continuously developing and perfecting over time, an ability that AI cannot yet replicate.

This new research not only helps to better understand language development in children, but also opens up new approaches for AI. Scientists believe that for computers to learn natural language like humans, AI needs to interact more with the real world, including movement, touch, observation and feedback.

Additionally, the study suggests that children’s language learning patterns may reflect how humans evolved communication hundreds of thousands of years ago. Language may have emerged not just to convey information, but also for social connection, play, and education.

New research technologies, such as eye-tracking devices, voice-analysis AI, and 3D brain models, are helping scientists gain deeper insights into how children learn and process language in real time.

As Professor Rowland shared: "If we want AI to learn language like humans, we need to redesign machines from the ground up, so that they not only process data, but also experience the world like children."

Back to topic

MINH HAI

Source: https://tuoitre.vn/ai-can-92-000-nam-moi-hoc-ngon-ngu-gioi-bang-mot-dua-tre-20250910100215411.htm