The Oral Logic by Xuan Ye
8 July 2020By Emily Fitzpatrick
In the early 2000’s, the Defense Advanced Research Projects Agency (DARPA), an agency under the United States Department of Defense working on social forms of Artificial Intelligence (AI), began to develop machine-learning agents that could cognitively engage with each other, their environment, and essentially ‘learn’ from their experiences in a simulation. During one simulation, two learning agents named Adam and Eve were programmed to know some things (how to eat), but not much else (what to eat). They were given an apple tree and were happy to eat the apples, but also made attempts to eat the entire tree. Another learning agent, Stan, was introduced and wanted to be affable, but eventually became the loner of the group. Given the natural development of the simulation—and a few bugs in the system—Adam and Eve began to associate Stan with food and one day took a bite out of him. Stan disappeared and thus became one of the first victims of virtual cannibalism. (1) Continue Reading