List of questions
Where should we avoid using Machine Learning in Financial Services?
Machine Learning is the hot topic at the moment however, are there areas within Financial Services that should be avoiding this technology?
How much of a real role will crypto currency play in Financial services in the future?
With the hype of Bitcoin and other crypto currencies there is the assumption that crypto currency is the future of Financial Services.
Q: What can we do to increase trust in artificial intelligence? Do you think that the key to building trust is transparency? And could this be achieved through regulatory rules around explainability?
With big data analytics there is always an element of human intervention - Data is an input and algorithms do what they are programmed to do. Ineffective oversight of these tools can lead to bias within decision making processes.
Q: Can the DLT/blockchain be trusted if it is decentralised and sits outside regulation?
The DLT/blockchain was designed so transactions are immutable, meaning they cannot be deleted. The ledger is updated through cryptography ensuring that the data can be distributed but not copied.
2161: FNZ currently have a Graduate Programme focused on one component of our organisation, Asset Services, from your experiences or insight, how do we effectively develop this further to include the other core areas of our business (Technology, Development, Product, Analyst...) to create an all encompassing FNZ Fintech Graduate Programme...? 2162: what do millenials want from an employer in financial services...
2161: 2 Year Graduate Programme including 6 month team rotations across asset services, completion of industry qualifications (CISI IOC) as well as 6 monthly performance criteria and continuous improvement presentations at each rotation. But this covers just one section of our business , how and over what time period is reasonable to expose them to the wider business for rounded understanding to develop our future leaders and top FNZ talent? 2162: To help further understand and maximise the requirements expected, working environment, opportunities and needs of this generation.
What will the future of business Credit Assessment look like?
Financial accounting APIs / Open Banking / Blockchain / Anti-fraud / artificial intelligence etc. means that more data than ever before is potentially available to assess the credit worthiness of businesses? But will it actually make a difference? Will it bring competitive advantage to the organisations that plough money into R&D to automate / improve credit decision making or will credit decisioning remain part of the realm of large credit reference agencies who will add further capabilities/services. Will default rates actually improve or will simply the cost of credit decisioning rise with new capabilities.
Given advances in technology, how can Monte Carlo simulations be scaled effectively?
Points to consider: - cloud computing, vectorized programming languages and modern hardware, for example, SIMD (single instruction multiple data) instructions and GPU (graphics processing unit) parallel computing; - how best to approach the processing of the data output, should modern big data techniques (Hadoop/MapReduce) be used.
How should audit/model validation/backtesting function in a data-driven modelling approach?
The "classic" modeling paradigm entails to data collection & cleaning -> model estimation -> bactesting/validation -> continuous monitoring. This is also codified in regulation. Data-driven/machine-learning-based modeling disrupts all elements of the chain above. How should the supervisory functions adopt to the new environment?
How should responsibility for modelling/business decisions be shared in a data-driven/AI-based decision making process?
How to quantify model risk? How to manage that the underlying data set is constantly changing? What level of transparency is achievable in a data-driven decision making process?
We would like to be able to measure effectiveness of multi-channel (SMS, Email, call, etc) promotional messages for different personal banking products depending on the product type, channels and their sequence. Considering a number of combinations of product times different channel sequences the problem grows exponentially. We would like to structure the problem to be able to answer these questions en mass rather than picking up a specific scenario to analyse. As we build a better understanding of the most effective combinations and sequence of channels to use, it would be great to explore what this tells us about how we can attribute the credit for each sale, across the various channels used.
What are the pitfalls that need to be considered when applying survival analysis in the domain of modelling customer defaults on a lending product (mortgage, loan, credit card)?