Generative Artificial Intelligence for Code Generation

In this slide deck we dive into generative artificial intelligence, in particular language models for code generation.  First we differentiate the use-cases research software engineering and data science. We learn how to identify training data of large language models in specific cases. Software Engineering Benchmarks show us how capable LLMs actually are and drive development in this domain. We dive into cyber-security and business aspects behind the technology. Finally, we learn how benchmarking in one specific scientific domain allows us to study capabilities of LLMs in more detail and how we can gain trust in AI-generated code, potentially without knowing what results it is supposed to generate.