Definitely the blend of articles. You need to decide what you don't include as what you do. We don't just want a list of links (you can get that anywhere) we want curation and your opinion!
All Apple have to do now is figure out how to squeeze an M4 ultra equiv complete with half a gig of VRAM/RAM into an iPhone - then they'll probably have something good to run most people's local AI needs (and can tweak and integrate an open source model - like they did with BSD for MacOS). Frankly that doesn't feel like an impossible task in the next 5(ish) years.
Great edition. One of the best Sunday's for awhile! Will dig into the 80K article in particular
Thanks Colin! What made it particularly valuable to you?
Definitely the blend of articles. You need to decide what you don't include as what you do. We don't just want a list of links (you can get that anywhere) we want curation and your opinion!
All Apple have to do now is figure out how to squeeze an M4 ultra equiv complete with half a gig of VRAM/RAM into an iPhone - then they'll probably have something good to run most people's local AI needs (and can tweak and integrate an open source model - like they did with BSD for MacOS). Frankly that doesn't feel like an impossible task in the next 5(ish) years.
The 80,000 hours article is gold.