The study found that artificial intelligence (AI) systems are unlikely to reach human cognition unless they are connected to the real world through robots and designed based on evolutionary principles. Cognition is the mental process of acquiring knowledge and understanding through thought, experience, and the senses.
Research published in the journal Science Robotics found that artificial intelligence systems would not resemble real brain processing, no matter how large their neural networks might be or the datasets used to train them, if they remained disembodied.
Researchers at the University of Sheffield in the UK noted that current artificial intelligence systems such as ChatGPT use large neural networks to solve difficult problems such as generating comprehensible written text.
These networks teach AI to process data in a way that is inspired by the human brain, as well as learn from its mistakes to improve and become more accurate.
Although these models resemble the human brain, the researchers said there are also important differences that prevent them from achieving biological-like intelligence. First, they said, real brains are embodied in the physical system of the human body, which directly senses and acts in the world.
Being embodied makes brain processes meaningful in ways not possible with disembodied AIs, which can learn to recognize and generate complex patterns in data but lack a direct connection to the physical world, the researchers said. Therefore, such AIs have no understanding or awareness of the world around them, they said.
Second, the human brain is composed of several subsystems that are organized in a specific configuration—known as an architecture—that is similar in all vertebrates from fish to humans, but not in AI.
The study suggests that biological intelligence—as in the human brain—evolved because of this specific architecture and how it used its connection to the real world to overcome challenges, learn and improve over the course of evolution.
This interaction between evolution and development is rarely considered in AI design, according to the researchers.
“ChatGPT and other large neural network models are an exciting development in AI, showing that really hard challenges like learning the structure of human language can be solved,” said Professor Tony Prescott, a professor at the University of Sheffield.
“However, these types of artificial intelligence systems are unlikely to progress to the point where they can fully think like the human brain if they continue to be designed using the same methods,” Prescott said.
AI systems are much more likely to develop human-like cognitive abilities if they are built on architectures that learn and improve in a similar way to the human brain, using its connection to the real world,” he added.
Read Now:Stephen Hawking’s famous theory could mean that the entire universe is doomed to black holes
