@Shekswess I think its because of the ability to fine tune your model (most importantly the size of uour choice) and be able to host it for inference tailored for your usecase. Oss 📈
@jxmnop This is very cool.
I am interested in finding out how close the distribution is between this and the original base model.
We can try this on an open source model for which we have both base and aligned versions to find that out.
@captmatsat I still like his lyrics in kannaana kanne
“கடல் அலை போல உன் கால் தொட்டு ஒரசி கடலுள்ள போறவன் நான் இல்லடி. கடல் மண்ணை போல உன் காலோடு ஒட்டி கரை தாண்டும் வரைக்கும் இருப்பேனடி”
But i cant find any such metaphors in dheema dheema
"Forget SkyNet robots from the future -- stop the shitheads in the present": https://t.co/ECsNujC1di
The truest take on AI's existential risks I've seen in awhile (featuring work from friends like @mmitchell_ai and @timnitGebru!)
sorry to be a spoiler but there is no way around a genuinely socially responsible AI without directly tackling systemic inequalities and building accountability mechanisms on the current concentration of power in the hands of big corps/AI companies