A writer who began his professional career in 1985 says working closely with large language models has changed his relationship with writing, turning AI from a tool for producing text into a participant in the development of ideas. The experience, he argues, can expand a person’s thinking while also making it easier to rely on a machine to do too much of the cognitive work.
In an essay for Psychology Today, he recalls adapting to earlier changes in writing technology, from pencil and paper to computers and internet search. Those tools altered how he wrote, he says, but remained mainly mechanical. Large language models were different: he could give one an incomplete idea and receive a response that shifted or advanced his thinking.
He describes the resulting exchange as iterative intelligence: a back-and-forth in which a person and an AI model develop an idea together. An unexpected suggestion can redirect the writer’s thoughts, making the process more than simply asking a system for an answer. He says this is why he questions descriptions of AI as merely a way to outsource intellectual work.
But the same exchange can blur who is doing the thinking. As a person responds to AI-generated suggestions, the original idea changes, and it may become difficult to identify where a thought came from. The writer calls this “thought losing its address,” particularly as AI systems become more capable of acting with less direct input.
He also questions whether removing the difficulty of writing is always an improvement. AI can help overcome the frustration of getting an idea onto the page, but he says the struggle of working through an unfinished thought may itself contribute to progress. If models shorten that journey, it is not yet clear what parts of the process are lost.
The writer connects this concern to what he calls “cognitive surrender”: people can take less part in the thinking even as the finished result improves. He uses “ghost thought” to describe cases where AI may supply not only the writing but also the underlying idea, leaving the person to approve a result that still feels like their own.
He remains enthusiastic about AI and says it has helped him think of things he might not otherwise have considered. At the same time, he acknowledges sometimes accepting its output too quickly. After years of working with the technology, he says he is still trying to understand where the tool ends and his own thinking begins.
Sources: Psychology Today.