Rich Hickey: Let AI Read and Code for You, and You Lose the Chance to Learn
Rich Hickey argues that delegating work to language models can cost you the chance to learn it yourself. Commenters pushed back, saying it depends on how you use them.
Rich Hickey posted a note on LinkedIn on October 12 about how we should think about AI. His argument: handing work over to a language model (an AI that reads and writes text) is not using a tool; it is automation.
The line Hickey draws between the two comes down to whether a person stays involved in the work. If you are the one doing the work, checking the results, and building skill in the process, that is tool use, he explained. A language model that takes an instruction and does the job on your behalf, by contrast, falls on the automation side. And if you hand off reading or coding, he argued, the learning you would have gotten from doing that work disappears along with it.
The comments brought pushback. One counterargument: a language model can also be a tool, depending on how you use it. The focus of the debate is not the AI label itself, but how much of the work a person chooses to delegate.
For anyone using AI to read and to code, the post leaves one question hanging. Do you choose to get the output quickly, or to learn by doing the work yourself? What difference the scope of delegation and the degree of human involvement actually make to learning is the open question going forward.