According to a new study by scientists from the Max Planck Institute of Linguistic Psychology in the Netherlands, in collaboration with the ESRC LuCiD Center in the UK, even the smartest machines cannot defeat beginners in language learning.
"While machines can process huge amounts of data at extremely fast speeds, children are far ahead of artificial intelligence in learning natural language," the institute said.
The main reason is that children use all senses such as vision, hearing, smell and touch to feel the world and build language skills.
Meanwhile, machines, including AI chatbot, mainly learn from written text passively instead of through a proactive process driven by increasingly developed social skills, awareness and movement. This means that AI misses the rich combination of images, sound, and other sensory signals like the way humans learn language.
According to scientists, children not only wait for the language to come to them but also proactively explore the surrounding environment, continuously creating new opportunities to learn. This is completely different from machinery.
When publishing their findings in the journal Trends In Cognitive Sciences, the researchers said: "If humans learned a language at the same rate as ChatGPT, it would take them 92,000 years."
Professor Caroline Rowland, lead author of the scientific research on this issue, emphasizes that all of these help children grasp language very quickly. "AI systems process data... but children actually live with it," she said.
According to her, AI researchers should understand how children learn language as a prerequisite to build better language models.
"AI researchers can learn a lot from children," said Rowland. If we want machine learning to be as good at human language, perhaps we need to rethink how to design them from the beginning.
