One weekend last fall Stanford CS doctoral student Andrej Karpathy decided to create a neural net that generated baby names. He fed in a list of 8,000 actual baby names, and the net iterated 8,000 results, 90% of which were new.
I literally grabbed a chunk at random, and then scrolled down until I had to stop. Jean-Xelly is great. Xelly would be a badass variation of Shelly, or when she's in her rebel futurist phase, the kid could go with Zelly.
Jatuddey
Climh
Stera
Xileh
Jonnie
Sankie
Rossell
Gie
Wolmonh
Aloria
Sveth
Cristam
Demas
Elwis
Aderic
Ilmah
Cedella
Jean-Xelly
Anyway, this name generating technique would work equally well for naming 8,000 MMORPG characters, or 8,000 tweet botnet accounts as for naming 8,000 babies of the future. The limit is just your imagination.
#RandomExperimentSundays: LSTM Baby Name Generator [AndrejKarpathy's g+, via @caseyg]
namesGenUnique.txt [cs.stanford.edu]
