Global AI MapHISTORY / 2017
The transformer
Attention-based sequence modeling removed recurrent structure from the core architecture.
What happened next
Transformers became a common architecture for large language models and later appeared in vision, scientific modeling and robot policies.