Artificial intelligence (AI) company Antropic claimed to have found a structure similar to the think..

Anthropic researchers have identified internal neural patterns within their Claude model that resemble the functional structures observed in human brains during specific cognitive tasks. By isolating these features, the team demonstrated that they could manipulate model behavior to alter how the AI processes particular concepts. This discovery offers a new method for interpreting the "black box" of neural networks, potentially leading to better safety controls and deeper insights into how large language models represent information internally.