Are we though? We haven't even started training 1.58b models, synesthetic data and "model gyms" are very promising and new architectures have been coming out t

Are we though? We haven't even started training 1.58b models, synesthetic data and "model gyms" are very promising and new architectures have been coming out that offer real benefits over transformers.