A recent study conducted by the University of Sheffield in the United Kingdom showed that AI systems are unlikely to achieve human-like cognition unless they are connected to the real world through robots and designed using evolutionary principles. While there has been ongoing discussion around AI reaching human-like cognitive levels, there is another aspect of this discussion that is worth exploring – the possibility that AI will develop a unique, silicon-based cognition that humans may not fully understand.
Today's AI systems, such as ChatGPT, utilize large neural networks to solve complex problems, such as generating understandable written text. These systems learn from their mistakes, improve over time and become more accurate. When it comes to data processing, they draw inspiration from the human brain, but it must be understood that their functions are still fundamentally different from the human brain.
The main difference is the nature of their existence. The human brain is embodied in a physical system that interacts directly with the world, but AI systems lack this direct connection. AI can learn to recognize and generate complex patterns in data, but without physical embodiment, AI misses out on understanding or awareness of the world around it.
Second, the human brain is made up of multiple subsystems, organized in a specific architecture that evolves over time to overcome challenges and learn from the real world. This interaction between evolution and development is rarely considered in the design of AI systems, leading to significant differences in their cognitive processes.
Despite these differences, AI has made significant progress in solving complex problems. As Professor Tony Prescott of the University of Sheffield puts it, "ChatGPT and other large neural network models are exciting developments in artificial intelligence that show that truly daunting challenges such as learning the structure of human language can be solved." ”
While it's clear that current AI systems are nowhere near thinking like the human brain, they may be on a completely different trajectory. What if artificial intelligence, instead of developing human-like cognition, is developing a new form of cognition, a silicon-based form of cognition that is unique to it? This silicon-based cognition will be rooted in the digital world, shaped by machine learning algorithms, characterized by the ability to process large amounts of data at speeds unmatched by the human brain.
Recent advances in the development of artificial intelligence systems for controlling robots indicate the potential of this unique form of cognition. A recurrent neural network model consisting of multiple feedback loops is being used to improve the adaptability of the robot. While they are far from real brains, these models mark an important step in the cognitive evolution of artificial intelligence.
It is important to consider that cognition of AI development may not be fully understood by us. Just as we struggle to understand the exact mechanisms of our own brains, AI's silicon-based cognition may remain a mystery, a form of intelligence that operates on principles beyond our current understanding.
Dr Stuart Wilson of the University of Sheffield believes the next breakthrough in AI will come from more closely simulating how real brains develop and evolve. However, as we continue to push the boundaries of AI, we may eventually create a form of cognition that is as foreign to us humans as we humans are to AI.
Therefore, as we explore the field of AI further, it is important to be open to the fact that the perception of AI may not be as good as that of our human cognition. Instead, it may develop a unique, silicon-based cognition that is fundamentally different from human cognition, but could be just as magical and complex.
Thesis: https://www.science.org/doi/10.1126/scirobotics.adg6014