Has anyone explored using one of these stupidly-large language models to compare/contrast between languages, and see if there's any hope of Chomsky's Universal Grammar living at a deep level?
It seems like they would be ideally suited...
Unus Nemo likes this.
Unus Nemo
in reply to Blaise • •@Blaise
Sounds like a great project for someone with the right hardware. I unfortunately do not have the hardware to train or utilize any of the extremely large models. My ceiling is around 16b for usage and training. Not very practical in this type of research. It would be really interesting to see the results of this research though.