Re:Roboticst at OSU uses performing arts to make robots more...
by sgillen on Friday August 10, 2018 @10:59AM
(#17733826)
Yes but being a human they already have a pretty good idea what might be funny for the audience they want to perfmorm for.
So basically I think a computer would need a lot more trial and error than the human does, and I’m not sure a good way to allow for that magnitude of trial and error without having some other computer that has a sense of humor be the judge.
Re:Roboticst at OSU uses performing arts to make robots more...
by Isamu on Friday August 10, 2018 @11:05AM
(#17733907)
No, you'd want to bootstrap the process with some kind of machine learning to categorize humor and non-humor. It would have poor accuracy, no doubt, but it would be better than random generation.
I agree that it would be less efficient than in humans but with learning it might be able to converge on a set of strategies, possibly "cheats" that are likely to produce laughs.
The reason is stand-up comedians have to develop their act empirically. They try out material that they think might be funny, and they try it out in front of an audience. They go through a process of constant trial and error. This is in contrast to what most audiences think happens.