To develop the next generation of AI, Trinity research looked at how babies learn

28 Jun 2022

Image: © Joeri/

New research noted three factors that help babies learn and proposed ways these principles can be adopted to improve machine learning.

Researchers at Trinity College Dublin have said we could boost machine learning and AI by drawing insight from how babies learn information.

Dr Lorijn Zaadnoordijk, a Marie Skłodowska-Curie research fellow at Trinity, said machine learning has been responsible for many AI developments over the past decade, with massive datasets being used to train neural network models.

However, Zaadnoordijk added that progress is stalling in many areas because these massive datasets have to be “painstakingly curated by humans”.

“But we know that learning can be done much more efficiently, because infants don’t learn this way. They learn by experiencing the world around them, sometimes by even seeing something just once.”

‘Through interdisciplinary research, babies can help unlock the next generation of AI’

Researchers in Ireland and the Netherlands examined the neuroscience and psychology of infant learning in a bid to help overcome the most pressing limitations of machine learning.

In their study, they identified three factors that help infants with their quality and speed of learning. The team said the information processing of babies is guided and constrained; they learn from diverse, multimodal inputs; and their input is shaped by development and active learning.

The researchers proposed that by adopting these factors, it could give rise to previously unseen performance levels in unsupervised machine learning.

“Artificial neural networks were in parts inspired by the brain,” said Prof Rhodri Cusack, director of Trinity’s Institute of Neuroscience.

“Similar to infants, they rely on learning, but current implementations are very different from human and animal learning. Through interdisciplinary research, babies can help unlock the next generation of AI.”

To achieve this, the team proposed that machines will need in-built preferences to shape their learning from the beginning. The machines will also need richer datasets and a developmental trajectory, where experiences and networks change as they ‘grow up’.

Dr Tarek R Besold, a researcher at TU Eindhoven in the Netherlands, said AI researchers often draw “metaphorical parallels” between their systems and the mental development of human children.

“It is high time to take these analogies more seriously and look at the rich knowledge of infant development from psychology and neuroscience, which may help us overcome the most pressing limitations of machine learning.”

10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news

Leigh Mc Gowran is a journalist with Silicon Republic