Transformer architecture introduced — 'Attention Is All You Need'

Google Brain researchers publish the Transformer paper, replacing RNNs with self-attention. This becomes the foundation for GPT, BERT, and virtually all modern large language models.