An API or JS SDK for Fine Tuning GPT-3 would be great!

My two cents: I think making a lower code solution entry point for the production of api connected AI models for aspiring young entrepreneurs’ various use cases is a potential killer product

Some dashboard to manually correct or verify the API past responses, to build a new curated dataset. This dataset can be then available for fine-tuning.

- Unstructured output. - Going overboard with unstructured input. Too many moving parts that exhaust tokens. - Ability to provide global tokens for a domain and incremental tokens for each question in that domain.

Need top_k param. Also if it possible share fine-tuning model to another organization

Can’t set a password besides Google login… makes it a pain to login while traveling

GPT-3 is an incredibly powerful tool but the docs and tutorials are very limited. You guys should be pumping out demos and tutorials to open up peoples' minds to the possibilities because they are 🤯

For us it was the cost.

Getting to playground is laborious. I beeline for playground whenever I have questions about life, the universe and everything

When I give it code to explain it tends to explain the outermost statements but fails to explain the inner statements