how can one avoid building a chatgpt wrapper?

Definitely the strongest approach is fine tuning / training your own model with proprietary data. Niche/unique data, stronger if it’s structured data. 

It’s getting easier and easier to fine tune a mode to specific tasks too.

Definitely the strongest approach is fine tuning / training your own model with proprietary data. Niche/unique data, stronger if it’s structured data. 

It’s getting easier and easier to fine tune a mode to specific tasks too. 

https://x.com/mattshumer_/status/1694412623477879201?s=46&t=kbUluqkrtSAIaGiJXuXVAw

FID 1580 | dad 2x. building SkillsEngine.com — product. design. art. code. philosophy. technology. economics. systems.

I think it’ll probably be a combo of all of those. Also I think we’ll see some existing companies rise up that have built unique data sets over the years but who previously didn’t have a way to leverage that data meaningfully.

i wonder to what degree will data become proprietary. will it be openai vs deepmind? us vs ccp? smaller startups against each other? or every individual against ourselves?