A stub.
In 1960 Licklider wrote Man-Computer Symbiosis. The argument was that the interesting future was not machines replacing human thinking, and not humans doing it all unaided, but a partnership where each side does what it is actually good at.
Sixty years later most of the writing has gone in a different direction. Either the machine takes over, or it does not. The partnership framing gets treated as a compromise position, which I think is a mistake.
What Licklider got right is that the division of labour matters more than the capability. Machines are good at recall, at scale, at not getting bored. Humans are good at knowing when something smells wrong, which is a genuinely hard thing to specify.
The place this gets concrete for me is identity. A machine can compare a million pairs of records without complaining. It cannot tell you that this particular pair matters more than the others because of something nobody wrote down.
What I want to think through
- Where the handoff should sit, and how a system knows it has reached it
- Whether “human in the loop” is a real design or a liability shield
- What Licklider assumed about computers that has since stopped being true
Related
Some of this connects to what an agent should refuse to do.