Then how can you confidently say we don't think 'like' Transformers/Attention/Statistical models/etc/etc?