And the problem is that since these dickwads want to close source their neural networks, we wasting billions of watts training on similar data sets repeatedly generating models with minor differences over and over again. If there was a truly awesome open source model, then we could train on top of it and create more complex and complex models targeted at the good of humankind. But no, stakeholders must get profits. When will they understand that the entire humanity is a stakeholder in this planet.

[–] yogthos@lemmy.ml 25 points 8 months ago (2 children)

Good news is that open source approach appears to be advancing faster than closed models https://steve-yegge.medium.com/were-gonna-need-a-bigger-moat-478a8df6a0d2

[–] Chinchillax@lemm.ee 6 points 8 months ago* (last edited 8 months ago) (1 children)

That article is from 10 months ago.
Is open source catching up still? Or have the goalposts moved too much since then?

[–] yogthos@lemmy.ml 10 points 8 months ago

Open source models are still advancing quickly, pretty much everything the article mentions is still valid today.