augmenting human intellect

gist

Ryo Lu argues that AI tools should extend what a person can understand and change, not only produce more software with fewer people. He traces that aim to Douglas Engelbart's 1962 report on augmenting human intellect, J.C.R. Licklider's 1960 paper on human-computer symbiosis, and Dan Ingalls's 1981 design principles for the Smalltalk programming system. Lu separates chores he is glad to hand off from hard work through which people find out what they mean. He asks for systems that people can reshape, question, undo and learn from, and says better models alone will not deliver them.

ideas

• Engelbart started from the person, not the machine. Lu reads the 1962 report as placing capability in the whole system of a person, their tools, their language and their methods.
• Licklider wanted computers to help form ideas. Lu draws on Licklider's 1960 distinction from systems in which the remaining people mainly serve the machine, and asks whether a tool helps you discover what you mean or makes you submit a request and wait.
• Not every difficulty is overhead. Lu wants search, file moves and repetitive work taken away, but says moving pieces around, seeing the wrong version and taking it apart can change what a person set out to make.
• Systems should fit how different people think. Citing Ingalls on personal mastery and on computer models that fit the human mind, Lu asks for systems people inhabit differently, through pictures, writing or simulation.
• Lu would judge AI by what newcomers can do. He asks what a person who could not participate before can now understand, express and change without asking permission.

quotes