Modern data on brain function don’t support the classic explanation of helplessness in babies. New research suggests that babies’ brains don’t mature as much as previously thought, but instead use their postnatal “helplessness” periods to develop basic patterns similar to those that drive generative artificial intelligence.

The study, led by a neuroscientist from Trinity College Dublin, was recently published in the journal. Trends in Cognitive Sciencefound for the first time that the classical explanation for infant helplessness was not supported by modern brain data.

Compared to many animals, humans are helpless for a long time after birth. Many animals, such as horses and chickens, can walk on their day of birth. This long period of helplessness exposes human babies to danger and places a great burden on the parents, but surprisingly they have withstood the evolutionary pressure.

Interspecies research articles

“Since the 1960s, scientists have believed that the helplessness exhibited by human infants is due to limitations at birth. It was believed that babies with large heads were born prematurely, leading to immature brains and a helpless period lasting up to a year. We wanted to find out why human infants remain helpless for such a long period,” explains Professor Rhodri Cusack, Professor of Cognitive Neuroscience and lead author of the paper.

The research team includes Professor Cusack, who measures the brain and mind development of babies using neuroimaging; Prof. Christine Charvet of Auburn University, who compared brain development across species; and DeepMind Senior Artificial Intelligence Researcher Dr. Marc’Aurelio Ranzato.

“Our study compared the brain development of different animal species. It was based on a long-term project Translation Time“This equates the relative ages of different species, suggesting that the human brain is more mature at birth than many other species,” says Professor Charvet.

The researchers used brain imaging and found that many systems in the baby’s brain were already functioning and processing a rich stream of information from the senses. This contradicts the long-held belief that many systems in the baby’s brain were too immature to function. The team then compared the training in humans to state-of-the-art machine learning models, in which deep neural networks benefit from a “desperate” pre-training period.

In the past, AI models were trained directly to perform the required tasks, such as teaching a self-driving car to recognize what it sees on the road. But now, models are first pre-trained to see patterns in large amounts of data before they perform any major tasks. The resulting base model is then used to study specific tasks. This has been found to lead to faster learning of new tasks and better performance.

Implications for the future development of artificial intelligence

“We propose that human babies similarly use the ‘helpless’ period in infancy for pre-training, learning strong foundational patterns that continue to support cognition later in life with high efficiency and rapid generalization. This is a feature of OpenAI’s ChatGPT, which has led to major breakthroughs in generative artificial intelligence in recent years.” It’s very similar to powerful machine learning models like si or Google’s Gemini,” Professor Cusack explained.

Researchers say future research on how children learn could inspire the next generation of artificial intelligence models.

“Despite major breakthroughs in AI, basic models consume enormous amounts of energy and require much more data than children. Understanding how children learn could inspire the next generation of AI models. The next step in the research will be to directly compare brain learning with AI,” he said.