Can a computer learn a language like a child? Researchers have argued that it can, and should. They think we need to change how artificial intelligence (AI) learns.
Children learn languages by talking with people around them. They play with words and try to guess what others mean. This helps them learn how to use language properly.
"This process, in which language is acquired through interaction and meaningful context, is at the core of human language acquisition," says Katrien Beuls in a Vrije Universiteit Brussel press release.
However, today's AI, like ChatGPT, learns differently. A large language model (LLM) looks at lots of text and sees which words often go together. This allows it to make texts that look like what humans write. But it has problems. It can make up things or show biases. It also struggles with thinking like humans and uses a lot of data and energy.
Making AI learn by interaction
The researchers suggest a new way where AI learns by talking with others in a real-world setting. They did experiments where AI agents learned words and phrases by interacting with their environment. This method makes the AI less likely to make mistakes or be biased. It also uses less data and energy, which is good for the environment. Plus, it understands the meaning and purpose behind words better, like humans do.
The researchers have described the methods and results of this study in a paper published in Computational Linguistics.
The researchers believe this approach could lead to better language AI that works more like human brains. They think it's an important step for making AI that understands and uses language in a more natural way.
"Integrating communicative and situated interactions into AI models is a crucial step in developing the next generation of language models," the researchers conclude. "This research offers a promising path toward language technologies that more closely resemble how humans understand and use language."