|
|
|
|
|
|
by mutkach
313 days ago
|
|
|
Not only "AGI" is cancelled but they also sort of admitted that so-called "scaling" "laws" don't work anymore. Scaling inference kinda still works, but obviously is bounded by context size and haystack-and-needle diminishing accuracy. So the promise of even steadily moving towards AGI is dubious at best. |
|
|