Transformers and Self-Attention Importance

Michael Agros explains older model limits and how transformers' parallelism and self-attention capture long-range context.