Summary

In this chapter, we covered the “dark side” of generative AI technologies, exposing its associated risks and biases, such as hallucinations, harmful content, and discrimination. To reduce and overcome those risks, we introduced the concept of Responsible AI, starting with a deep dive into the technical approach we can have while developing LLM-powered applications; we covered the different levels of risk mitigation – model, metaprompt, and UX – and then moved on to the broader topic of institutional regulations. In this context, we examined the advancements that have been carried out by governments in the last year, with a focus on the AI Act.

Responsible AI is an evolving field of research, and it definitely has an interdisciplinary flavor. There will probably be an acceleration at the regulation level to address it in the near future.

In the next and final chapter, we are going to cover all the emerging trends and innovations happening...