The dramatic progress of recent years has resulted largely from increases in data availability and processing power, rather than advances in the fundamental theoretical foundations of artificial intelligence. If these foundations can be developed through targeted research, we will gain an understanding of what is missing from the current paradigm, and how it can be improved and its applications expanded — safely, and with human needs at the focus.
A stronger theoretical basis may also help us solve problems created by the current nature of AI. The field of explainable AI is aiming to create a better understanding of how ML algorithms work, with increased reliability and transparency.11 This will have an important impact on applications, as it will then be possible to deploy AI techniques in sensitive domains where liability is paramount (for example, the health, financial, legal and engineering spheres).12
