Review of The Lifecycle of Software Objects (2010)

Text

Ted Chiang (writer).

Read in 2018.

Pretty dry. It’s tempting to read Ana’s conclusion, i.e. that experience makes the mind, as a premise. It is surely a fallacy. The same year this novella was written, DeepMind Technologies was founded. They eventually produced AlphaGo and other AIs that achieve a non-general superintelligence with no real experience beyond playing against themselves. The novella raises no concerns that would prevent this approach from boosting the development of a “digient” to a human-equivalent level. It seems simply wasteful to emulate a biological mind to the point of introducing major flaws in memory and deductive logic around such an abstract core that it leaves open the possibility that the digients could eventually gain control over their own reward systems, i.e. set their own emotional priorities. Compared to “Understand” (1991), this story is less mechanistic and weaker for it. Compare also “Mirror Girl” (1998), where unusually devoted human interaction “raises” an AI to a useful level.

The use of Lojban is a nice detail. The language, or rather a horrific relaxed version of it, has been seriously proposed “for communication between humans and AGIs” by Ben Goertzel.

References here: “Words We Say Instead” (2020), The Creator (2023).

text fiction cyberpunk