Last week at the Deep Learning in Finance Summit in London, we were joined by some of the leading minds working in AI to disrupt the financial sector. With speakers from Barclays, LV=, Prudential, HSBC and many more, we learned how the most cutting edge deep learning technologies are being applied in industry.
I spoke with Adam McMurchie, who is currently working in the digital innovations studio at RBS. He's focusing on rapid prototyping of new technologies, driving automation and conducting POCs of new tools, and then evaluating them to leverage the benefits for the bank. He explained that ultimately his main goal 'is to future proof technologies & services that run the bank.'
Upon asking Adam how he started his work in Finance, Ai and DevOps, he explained that all of these disciplines are intrinsically inked in a big way. 'DevOps was first unveiled in 2008 as a reaction to outdated software development practices. It was embodied with a sense of urgency during the financial crash. Since then DevOps has become synonymous with automation and the latest innovations in finance such as smart banking assistants that leverage both A.I & DevOps capabilities.'
Adam's entry into the field grew from his long living passion about green field technology, as finance is undoubtedly one of the best platforms for putting theory into practice.'This is because unlike other specialist domains, finance requires broad spectrum use of both horizontal and vertical A.I (solving general & niche problems respectively). A.I for A.I’s sake however is one of the most commonly adopted strategies in big business, this is what drove me to specialise in these domains, as I could see the need for best practices and raising A.I awareness in the higher echelons of finance.'
I went on to ask more about deep learning in finance, as well as RBS's current applications of these technologies:
What challenges are currently being faced in banking, and how is deep learning helping to solve them?
With the introduction of new reforms such as PSD2 and the growth of digital platforms, there are increasingly more new ways to bank and transact. Ultimately this means more data to work with and more scenarios to take into account when considering things like fraud and processing loans. To illustrate by example, many traditional fraud model providers require weeks or even months of reworking when there is a change. Even if it is a small change such as adding a new customer token field to the database.
Deep learning (DL) has become a powerful tool in addressing these problems, firstly there are a variety of DL models to choose from which allows for fine grain control over the adjustment period and predictability. For example LSTM neural networks are ideal for pattern recognition (such as establishing a customer spend profile) but responds well to change and doesn’t require a rebuild of the model from scratch. Additionally deep learning makes counter intuitive predictions, this is the corner stone of its success as it remediates blindspots, i.e. where loan models fail to spot vulnerabilities and fraud models let malicious activity slip past.
Ethics is currently a hot topic in AI, and in finance, there's a huge amount of sensitive data. What are your thoughts on the best practices around this?
To problem is best tackled by addressing the root cause, namely the lack of on site technical expertise. You can’t remediate A.I related data risks if you don’t have A.I specialists.
Firstly it is absolutely imperative that financial companies realise their responsibility for developing their own pool of A.I specialists. This does not mean hiring several PHD graduates and relegating them to a small think tank for passive consulting. Quite the opposite, A.I technologies are emerging full spectrum, naturally this means that the A.I specialists need to come from diverse backgrounds. A robust team should have strong analytical professionals who can produce models, bolstered by technical domain specialists who know how to prioritise and leverage the technology effectively, finally, guided by a product owner with a business vision and a A.I roadmap. Essentially this is a fully functional A.I agile team, a strong team will become self perpetuating in terms of added value.
This gets the company a long way in addressing the problem regardless if they are developing the tech in house or working with an A.I partner. It is important to note however, every time a bank enlist a partner to help with data exploration, a layer of risk is added. This is a two way street, in that small finTechs providing niche analytical services for the banks need to prove themselves and the banks need to maintain their core A.I in-house resource in order to adapt rapidly to change as not to risk falling behind in the A.I services arms race.
How is software and AI transforming the banking industry? Are people more cautious, and if so, should they be?
A.I is creating serious waves in the banking industry at present and is as much a cultural change as it is technical. That said, the impact of A.I in the workplace for banking is no different from any other tech industry; in that people are initially skeptical and fearful of change, especially when their core duties are at risk of being automated. In practice however, it isn’t as simple as that, affected employees will generally play a central role in setting up the automation, even if they aren’t proficient in A.I themselves. Many people I have worked with actually feel a sense of empowerment and as a result and become proficient in A.I themselves albeit at a high level. This however still adds significant value. Those employees can then use their experiences as a spring board into more sophisticated roles.
From the technical side we are witnessing an A.I chatbot arms race where banks are working hard on producing more intelligent automated interactions with both customers and staff. This isn’t just automation for automations sake, it effectively boosts the customer service availability from X% to 100% in the areas where it is proficient.
With smart monitoring, we are gaining meaningful insights into both customer activity and our own technology stacks. There are a range of intelligent tools out there which can monitor the entire bank network and provide rapid response info & alerting on malfunctioning nodes, servers, databases and more. This is essentially automating a huge chunk of service availability duties. As with all good automation, A.I is not only assuming those tasks but outperforming humans by several orders of magnitude.
Which other industries are you most excited so see implementing AI for a positive impact in the next 5 years?
A.I startups are forming every week, apart the usual media focus of autonomous driving and smart assistants, the A.I community is bubbling with activity. The next five years will see a growth of A.I akin to the spread of the internet in the late 90s. The ‘Is your company online yet?’ marketing drive could well see a reboot with the term A.I
Right now the hottest area in A.I with young rockstars has to be block chain, cryptocurrency and contract languages such as solidity. This activity hints at the creation of a new platform for democratisation of A.I (by making it decentralised). This is really important for the long term prosperity of A.I, I think this is an area where we will even see thebirthplace of AGI - Artificial General Intelligence, an A.I that can “learn to learn”.
For me I am quite excited to see where new not so obvious but equally novel applications of A.I will take place. There are companies such as Tetra that are leveraging advances in speech recognition to generate detailed notes from your phone calls. Hyper science is automating remedial office work and is able to extract data from forms easily using optical character recognition. JetLore uses consumer behaviour as input to a model and is able to output structured data.
Finally the most impactful use for A.I in general is the application in fighting climate change, from young high school students who develop a robot for sorting recycled trash to dynamic power algorithms which google have already implemented with deep learning to save hundreds of millions of dollars in energy monitoring.
Simply put in five years time, most of what I have covered will be apparent to the public, and become part of high school curriculum. All major industries from farming to even the arts will use A.I in some way. The question of who will be the leaders in this field remains to be seen.
Couldn't make it to the Deep Learning in FInance Summit but keen to learn more? Join us in New York this September for the AI in Finance Summit to learn from global leaders in the space.