Heavy human workloads and prospects for explosive AI self-improvement

One post questions whether AI has entered recursive self-improvement, pointing to the intense, often narrow work people still do at frontier AI labs. The author expects human effort to progressively decrease in that scenario and suggests AI may be approaching it rather than already being there. They see research automation accelerating progress, but not in an immediately explosive way. A reply adds that even automating all the digital parts of research would leave human interactions and relationships as sources of friction.