Grounded language acquisition through the eyes and ears of a single child. (https://pubmed.ncbi.nlm.nih.gov/38300999/)

These scientists wanted to understand how babies learn their first words by connecting what they hear with what they see. They studied one child from when they were 6 months old until they were 25 months old.

To do this, they used a special camera that the child wore on their head. This camera recorded everything the child saw and heard. The scientists then used this recorded information to train a special computer program called a neural network.

The neural network learned to match the words the child heard with the things the child saw. It learned to connect the sounds of words with the objects they represent. The scientists trained the neural network with 61 hours of recorded data, which is a lot of information!

The amazing thing is that this neural network was able to learn many word-meaning connections just like the child did. It could understand what different words meant by looking at the objects in the recordings. It even learned to understand new things it had never seen before, just by using the words it already knew.

This study shows that babies can learn a lot about words and their meanings by seeing and hearing things around them. The neural network learned in a similar way, using the information it got from the child's experiences.

Vong WK., Wang W., Orhan AE., Lake BM. Grounded language acquisition through the eyes and ears of a single child. Science. 2024 Feb 2;383(6682):504-511. doi: 10.1126/science.adi1374. Epub 2024 Feb 1.