The talk is about OpenAI's GitHub copilot. Based on an encoder-decoder model, the GPT-3 Transformer implements a tool for speech processing. With the help of an attention mechanism, the GPT-3 Transformer creates natural language texts. However, GitHub Copilot is more than auto-completion of recognized speech patterns. It uses the provided context to synthesize a suitable code. Together with the Model Codex, which is built by Guthub and OpenAI, it suggests numerous possibilities of suitable programming snippets to a natural language input.