If anything, the transformer style attention architecture and language models have pushed AGI further into the future because the vast majority of finding for AGI research is going into the dead end of transformer style architecture and language models, pulling it away from methods that actually could achieve AGI.