Gregory's Server@tinker Isn't the idea though that it's a big up front cost so that it's cheaper over time? We've spent a lot of resources training some huge models but a lot of the newer tech is building off of those foundations e.g. with transfer learning. I think there are certainly a lot of concerns to be had over (mis-)applications of machine learning, but this feels kind of like saying "we can't build a new solar panel every day! it's much more efficient to just burn wood"
|
2 comments
 @tinker I agree that some of the most publicized uses of "AI" are pretty pointless. But I see a lot more uses of machine learning in research, ideation, and automation that are huge quality of life improvements. For example: photographers using generative fill to remove objects from backgrounds, coders using it to draft docstrings, people auto-captioning for accessibility. I actually think those make up 80% of the uses and 20% of the spotlight on the tech. I see those costs continuing to improve |
@janeadams - Machine Learning on its own may have good applications. But the current iterations and goals of various ML are trash. Stealing words and art from people to spit out low quality derivations filled with error and falsehoods.
And all the while burning a metric fuckton of energy.
Automation is good. ML may assist in that. But it is not currently being applied to ease drudgery, is it?