Scientists Are Discovering Why Robots Can Never Learn as Well as Human Babies

Futuristic head with DNA strands and code suggesting artificial intelligence AI.

Artificial intelligence (AI) seems a threat to all facets of human life, from self-driving cars to sex robots, from automated check-out counters to AI religions. Even as robots become more sophisticated, however, scientists have been struck by the creativity and effectiveness of human intelligence. Those who started out in an attempt to make robots more human or use AI to boost human intelligence have come away with a sense of awe at the miracle of the mind — even in infants barely a year old.

Deb Roy, an AI and robotics expert at the Massachusetts Institute of Technology (MIT), joined his wife Rubal Patel, a speech and language specialist at Northwestern University, in setting up cameras to capture every moment of their baby son's life in 2005. At the time, Roy had boasted that he had already created a robot that was learning the same way children learn. He was convinced that if the robot got the same sort of input, it could learn just like a human child.

Years later, Roy told The Guardian's Alex Beard that the joys of fatherhood disabused him of these notions, giving him a deep appreciation for the uniqueness of the human mind.

"The first time his son uttered something that wasn’t just babble, Roy was sitting with him looking at pictures. 'He said ‘fah’,' Roy explained, 'but he was actually clearly referring to a fish on the wall that we were both looking at. The way I knew it was not just coincidence was that right after he looked at it and said it, he turned to me. And he had this kind of look, like a cartoon lightbulb going off – an ‘Ah, now I get it’ kind of look. He’s not even a year old, but there’s a conscious being, in the sense of being self-reflective,'" Beard wrote.

“I guess, putting on my AI hat, it was a humbling lesson,” Roy exclaimed. “A lesson of like, holy sh*t, there’s a lot more here.”

"Roy was no longer sure you could bring a robot up like a real human – or that we should even try," Beard noted. "It didn’t seem there was much to gain by developing robots that took exactly one human childhood to become exactly like one young adult human. That’s what people did."

Even this startling conclusion barely scratched the surface of human invention. The Guardian reporter noted that all this uniqueness in human learning came "before you got into imagination or emotions, identity or love – things that were impossible for [the robot intelligence] Toco."

In describing events with his son, the AI researcher said he "had been blown away by 'the incredible sophistication of what a language learner in the flesh actually looks like and does'. Infant humans didn’t only regurgitate; they created, made new meaning, shared feelings."

Roy quoted the autobiography of Helen Keller, the girl who was struck deaf and blind in infancy but eventually learned how to communicate starting at seven years old.

“Suddenly I felt a misty consciousness as of something forgotten,” she wrote, “a thrill of returning thought; and somehow the mystery of language was revealed to me. I knew then that ‘w-a-t-e-r’ meant the wonderful cool something that was flowing over my hand. That living word awakened my soul, gave it light, hope, joy, set it free! Everything had a name, and each name gave birth to a new thought. As we returned to the house every object which I touched seemed to quiver with life.”

This miraculous human learning powerfully contrasts with artificial intelligence. "For a robot, the acquisition of language was abstract and formulaic. For us, it was embodied, emotive, subjective, quivering with life," Beard wrote, explaining Roy's epiphany. "The future of intelligence wouldn't be found in our machines, but in the development of our own minds."

These grandiose claims did not come from a dogmatic belief in the uniqueness of human intelligence, but from hard-won scientific discovery into how to make humans smarter, using AI techniques. Sometimes the latest science bolsters the oldest religious truths — like the idea of people being uniquely intelligent, made in the image of a supremely intelligent God.

Ironically, scientists came to these conclusions after years of wrestling with learning. In 1995, researchers Betty Hart and Todd Risley had studied 42 Kansas City families and reported that "the number of words a child heard by their third birthday strongly predicted academic success aged nine."

Teachers had often thought that human learning involved mere data acquisition — learning words and concepts and storing them by the millions. This simplistic theory proved false, because the human mind is essentially social and driven by the desire for meaning, something more powerful than naked information.

Kathy Hirsch-Pasek, professor of early-childhood development at Temple University in Pennsylvania, wrote that "just as the fast food industry fills us with empty calories, what we call the 'learning industry' has convinced many among us that the memorisation of content is all that is needed for learning success and joyful lives." The title of her book tells the story: "Einstein Never Used Flashcards: How Children Really Learn and Why They Need to Play More and Memorize Less."

Even at the age of eight months, babies already understand things that computers cannot grasp. "They know that if I drop this plate on the table, the plate won't go through the table. That's amazing. They know that if I'm sitting across from you, and you can't see the bottom part of my body, I still have one," Hirsh-Pasek told The Guardian. These may seem like simple lessons, but they mark an essential difference between babies and AI.

Babies begin learning even before birth, The Guardian's Alex Beard reported. "Even in utero, babies are learning. At that stage, they pick up sounds," he wrote. "One-hour-olds can distinguish their mother's voice from another person's. They arrive in the world with a brain primed to learn through sensory stimulation."

"We are natural-born explorers, ready made for scientific inquiry," Beard quipped. This poetic statement echoed the best available science.

"We enter the world ready to 'read the perfect cues out of the environment,'" Hirsh-Pasek said. "We arrive ready to interact with other humans and our culture."

"The real genius of human babies is not simply that they learn from the environment — other animals can do that," Beard wrote. "Human babies can understand the people around them and, specifically, interpret their intentions."

Angela Prodger, co-director at the Pen Green Early Childhood Centre in England, told The Guardian that children are constantly communicating with adults, even as infants. Prodger argued that the social aspect proves far more important to learning than mere information.

"Creative play is the foundation on which creativity, language, maths and science are built. If you start too early with flashcards, you lose this developmental stage," Beard explained. "It's about being free. It's about risk-taking," Prodger emphasized.

Prodger argued that nurseries should be as social as possible. "If we're not addressing personal, social, emotional development first, you're not ready to learn," she said.

Hirsh-Pasek agreed. Machines can help children learn, but they only augment the most important type of education. "What the machine can't do is be a partner," Hirsh-Pasek told The Guardian. "It isn't social. It's interactive without being adaptive."

These discoveries represent a revolution in the approach to childhood learning. "Policymakers and laymen had twisted the science to fit their own ends," Beard wrote. "No scientist thought flashcards worked. No scientist believed you should start learning to read and write at an ever younger age. It was a fantasy of governments."

Beard explained the 2003 experiment from psychologist Patricia Kuhl, who tried to teach American infants Mandarin Chinese. Only the kids taught by a flesh-and-blood tutor, rather than audio or video aids, learned anything.

Even teaching early reading and writing can have negative impacts on children, according to a Cambridge study. New Zealand kids who learned literacy at age five fared no better than those who learned at age seven — when it came to reading ability at age eleven. Even so, "the children who started at five developed less-positive attitudes to reading, and showed poorer test comprehension than those who started later."

"These findings are clear: if you start on the decoding before you have an underlying understanding of story, experience, sensation and emotion, then you become a worse reader," Beard explained. "And you like it less. Treat kids like robots during early learning and you put them off for life."

Social learning sets humans apart from machines, and makes even the knowledge of infants a wonder to behold. These discoveries line up perfectly with a biblical understanding of humanity: according to Christian doctrine, God Himself is social in a way people cannot understand. God is the Creator of the universe, and the laws of physics radiate from His rationality. Made in the image of God, humans learn socially and have the key to decode the universe.

According to the Bible, human intelligence is literally a miracle. Beard attempted to explain it from an evolutionary standpoint. "Our long period of immaturity is a risky evolutionary strategy, making us vulnerable early on to predators or sickness, and delaying for many years our capacity to reproduce, but the payoff is immense," he wrote.

"We have evolved to be a species of teachers and learners," Beard explained. "Our ability to understand other people arrives around the ninth month, at a moment in their development at which babies begin to check the attention of others by holding or pointing at objects."

"Shared attention is the starting point of conscious human learning ... That's why it matters that we talk to our children. It's also why we can't learn from robots — yet," Beard quipped.

None of these scientific studies have fully explained the nature of human intelligence, but they have revealed that AI is not even in the same category as a baby's capacity to discover language.

The theory of evolution may explain the development of human knowledge — if so, it has a long way to go — but even if it does, this unique learning style will still seem nothing short of miraculous.

Science seems to confirm deep truths about humanity revealed thousands of years ago in the Bible, and these discoveries provide a firm reason to hope that AI will not replace human beings, even if it renders a few menial jobs obsolete.