I recall there was some noise that in 2026 Western open source will overtake Chinese one. Not seeing it yet. Tbh I plainly don't see it happening at this regulatory trajectory. If we get *closed* models in the Fable class *removed*, why permit a GLM 5.2 weights release?
No Digg Deeper questions have been answered for this story yet.
Like is the idea here that only very few actors can post-train GLM 5.2 to do some Fable Level Damage so it's OK? Or that the USG doesn't understand post-training? Or that they're blocking broad general capability as such, because they just value humans being retarded?
I recall there was some noise that in 2026 Western open source will overtake Chinese one. Not seeing it yet. Tbh I plainly don't see it happening at this regulatory trajectory. If we get *closed* models in the Fable class *removed*, why permit a GLM 5.2 weights release?
@teortaxesTex 监管这套逻辑确实有点左右互搏的意思
@teortaxesTex if you can post-train GLM 5.2 to do so, you can do the same post-train to kimi and even deepseek-v3 which was released 2 years ago.