In 2016, AlphaGo, developed by DeepMind, made a move in a Go match that stunned human players and commentators. The move, which seemed illogical, ultimately led to AlphaGo's victory over Lee Sedol, one of the greatest Go players of all time.
AlphaGo's ability to choose move 37 was not based on intuition alone but on its search machinery, which evaluated the future consequences of each possible move. This system, which constructed and searched a game tree with thousands of branches, represented a machine analogue of human System 1 and System 2 thinking.
The move was initially seen as a flash of machine intuition, but it was actually the result of AlphaGo's reasoning capabilities. Unlike today's AI models, which rely on pattern recognition, AlphaGo's architecture allowed for a structured, deliberative approach to problem-solving.
"I thought AlphaGo was based on probability calculation and that it was merely a machine," Lee Sedol said after the match. "But when I saw this move, I changed my mind. Surely, AlphaGo is creative."
The development of AlphaGo's reasoning model has implications for future AI systems, especially in fields like science and medicine, where trustworthy results and novel insights are crucial. Current large language models lack the structured reasoning capabilities that AlphaGo demonstrated.
DeepMind's research highlights the need for a new approach to machine reasoning, one that incorporates an epistemic state to represent what the system knows and how it manipulates that knowledge. This approach could lead to AI systems that produce knowledge that can withstand scrutiny.
Source: mittr
