dummyai

I found the model is good at capturing the statistics of the underlying text - it's quite good at producing stories that "sound like the original books". However, coherent reasoning is still in such models.

One example is "... She is a man....". These outputs clearly breaks proper reasoning constraints (and common sense)

Experiment framework seems a big need in research community. However, focusing on scalability rather than reproduciblilty might be a better strategy.

Reproduciblilty often comes naturally with scalability.

Thanks for taking time to sign up. We'd like to gauge interest in the research community first, and will be in touch once the mvp is released.

Sorry if it disappoints you.