In 2017, AI saw a breakthrough with transformers enhancing neural machine translation. OpenAI's GPT series progressed from GPT-1 with 117 million parameters using masked self-attention and Adam optimization, to GPT-4 with 1.8 trillion parameters and 120 layers. This rapid advancement has necessitated substantial computational power and led to regulatory measures, including a 2023 U.S. Executive Order imposing strict controls on AI data and compute resources.
















