AI, Machine Learning & Deep Learning Risk Management & Controls: Beyond Deep Learning and Generative Adversarial Networks: Model Risk Management in AI, Machine Learning & Deep Learning: Princeton Presentations in AI-ML Risk Management & Control Systems (Presentation Slides)

Abstract

The European Parliament Think Tank's Research Policy document 'Should we fear artificial intelligence?' reflects the ongoing mainstream debate between the Utopian and Dystopian aspects of AI and Machine Learning. "Powerful AIs can in principle be given nearly any goal, which is a source of both risk and opportunity. There are myriad possible malicious uses of AI and many ways in which it might be used in a harmful manner unintentionally, such as with algorithmic bias. Perhaps most fundamentally, the control problem will have to be addressed – that is, we will need to learn how to ensure that AI systems achieve the goals we want them to without causing harm during their learning process, misinterpreting what is desired of them, or resisting human control." Third in the series of the Princeton Presentations on AI and Machine Learning Risk Management & Control Systems, the current presentation develops fundamental guidance on the design, development, and implementation of AI, Machine Learning, and Deep Learning Models and Methods.

The 2018 Princeton presentation will focus on "the control problem" which is a critical prerequisite for AI systems to have positive impacts by further developing upon my prior two presentations that pioneered Cyber-Finance-Trust™ Model Risk Management & Model Risk Arbitrage™ practices at prior Princeton Quant Trading Conferences. Starting with the first technical report on the Bitcoin Blockchain Cryptographic Proof of Work; spanning latest developments in AI, Machine, Learning, Deep Learning, and, Generative Adversarial Networks; and, hedge fund algorithmic trading, the presentation generates interesting insights about the most critical role of risk management controls. Such role of risk management controls is most critical in not only getting the best out of AI, but also ensuring that the worst fears about the AI do not really come true.