We live at a time of exciting growth in development and deployment of AI in many sectors including manufacturing, financial services, policing, health and government. AI systems are increasingly easy to develop through the use of machine learning, and have the potential to provide remarkable economic and social benefits. However, increased use of AI is accompanied by new risks that we may not be ready to deal with. These include:-

• How to validate AI systems to ensure that they are safe, reliable and trustworthy and that they remain so over time.

• How to ensure that AI decision making and decision support will reflect the ethical and legal framework of our society. 

• How to ensure that AI systems are able to produce unbiased decisions that do not lead to discrimination. 

• Given the complexity of AI systems, how to test and verify them to check that they are functioning properly according to our expectations.

• How to ensure that AI decision making does not have detrimental side effects on the systems they are working in, or more generally in society. 

• How to explain how complex AI systems are behaving and making their decisions. 

• How to have confidence that AI systems will behave appropriately in unusual circumstances that they may not be familiar with. 

The Validate AI conference has been organized to address these emerging problems. The emphasis of the conference is on technical approaches to dealing with issues around AI validation and on how such developments may impact ethical concerns and society. Prominent speakers with practical experience of deploying AI in multiple sectors, including from the Alan Turing Institute, University of Oxford, Imperial College London, Google DeepMind, Capital One, FiveAI, and Her Majesty’s Revenue and Customs, will discuss relevant issues around validation and maintenance of AI. It will be an excellent opportunity for practitioners from all sectors to share ideas and concerns, coming together with the AI research community to promote collaboration. The conference will also act as a catalyst for decision makers to adopt more rigorous validation methods, as well as generate new areas of academic research in AI validation.