Member-only story
Can AI Keep Learning Without Us? — Autonomous Discovery, Direct Observation and Copyright
A dynamic mathematical model of AI’s changing dependence on human writing, solutions and observations
Suppose an AI system is asked why a machine has failed. It may find the answer in a repair manual, infer it from a circuit diagram, or request measurements that distinguish several possible faults. The repair manual records the work that has been already done by people, the inference capability uses knowledge that the system has already acquired, and measurements may supply information about this particular machine that neither the manual nor a more capable reasoner could establish without physical access to the machine.
These separate routes might support the same answer; however, they imply different sets of dependencies. A system that becomes less reliant on manuals, for instance, may remain reliant on a technician who takes measurements, whereas a robot could replace part of that work without supplying its own electricity, maintaining its laboratory, or deciding whose machines deserve attention.
To assess intellectual independence, we would ask whether the system could continue to acquire useful knowledge if…
