Content
@
0 reply
0 recast
0 reaction
Alessandro
@azeni
how can one avoid building a chatgpt wrapper?
5 replies
0 recast
1 reaction
Alessandro
@azeni
a key point might be proprietary data.
1 reply
0 recast
0 reaction
Max Miner
@mxmnr
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
1 reply
0 recast
1 reaction
Alessandro
@azeni
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?
1 reply
0 recast
1 reaction
Max Miner
@mxmnr
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.
1 reply
0 recast
1 reaction
Alessandro
@azeni
indeed. years in the internet harvesting data will have an interesting payout soon. i also wonder what non-obvious company will rise up with such obscure datasets. can be anything from reddit to discord to even uber or spotify. twitter made an interesting move sunsetting it's api. the internet might close again.
0 reply
0 recast
0 reaction
