CIPR members receive 5 CPD points and PRCA members receive 10 CPD points for listening to this podcast if they log it at their respective CPD programmes.
Russell Goldsmith spoke to a number of the speakers:
For more information from CFA UK, visit www.cfauk.org
Mikey explained that Kensho is a five year old Fintech company out of Cambridge, Massachusetts with a mission to make sense of the ‘messy world’ of financial data, through machine learning and analytics. He said that the company isn’t based completely on traditional venture capital. Their investors are a lot of big global banks like Goldman Sachs and Bank of America that have helped incubate the company and really understand the space, but they were acquired last year by S&P Global.
His session was titled ‘Machine learning vs. Economics’ and he described two ways that people like to approach problems.
He said that these seem at loggerheads. They seem like they don’t really get along and that you have to do either one approach or the other. From what he’s seen in finance, Mikey thinks we’re really heavily biased toward using the econometrics approach, where money is on the line in a heavily regulated industry as we really need that sort of explainability. However, in his session, he wanted to try to convince people that this duality is a false one as he believes that most of the good problems lie somewhere in between those two approaches, and if most of the good problems are there, then all of the good solutions to these problems are there too and it’s really about both sides easing their constraints and learning to approach problems a little bit differently. He said that the mark of a good problem solver is someone who knows how to how to walk that spectrum between machine learning and economics.
Mikey said there are quite a few typical problem examples that can help explain it and a lot of them in his experience have to do with teaching computers to understand language or so called natural language processing. Most recently, Kensho did one targeting automatic ingestion of data. He said , basically, you can think ‘I’m a company. I go and I buy a new data set and now I need to merge it into my existing data’. This is usually a very manual process because there are no standards on the way data is represented and so typically, it can take years to integrate a new data set. However, with some sort of machine learning techniques but also borrowing parts from the traditional econometrics approach, Kensho were able to take the time it takes to ingest a new data set from about a year to about a week. He added that as a financial services company, that’s huge as it lets them get data to the market much much faster than their competitors and eliminate roadblocks and get data into their product.
Mikey therefore wants people to understand that there is not really a duality between machine learning and economics and that if you if you relax some of the constraints that you may have ingrained in your mind, either if you’re a pure machine learning person or you are a pure econometrics person, you’ll really have access to being able to solve a lot of new really cool problems
To find out more information on this topic, visit www.kensho.com where there is a blog and supporting materials
Vinay hosted a discussion at the forum, looking at a new way of communicating with financial consumers, which focused on an advice engine that Envizage has created this for financial service firms. He said that the company recognised that there was a very significant engagement gap amongst consumers with respect to financial products and services and whilst there were many slick digital journeys, they were invariably journeys that took the consumer on the last mile of a very long and very complicated journey. These were what he calls action journeys. According to Vinay, they were invariably focused on consumers that were highly confident and had a high intent to purchase or to act. However, he said, that sadly, the data shows that most people don’t fit that brief. Envizage therefore felt that there was a very good use case to create an engine that could power digital experiences, very different in nature from client to client and country to country, that could help bridge this engagement and comprehension gap for the end consumer, that could effectively add context to financial products and put them in the context of the life, the future and the outcomes that most people understand and can embrace.
He added that what’s unique about Envizage’s engine, is that it actually cuts across saving and investing, protecting and insurance as well as borrowing.
Vinay explained that Envizage use very high quality data that often comes to them from their clients to simulate the future of an individual or a household. That future is, by nature, uncertain and there are a number of different sources of that uncertainty. It could be the ups and downs of the market or a member of the family falling ill or potentially even dying. What Envizage therefore do is use available data to simulate all of these scenarios with their realistic chances of happening to the households and at the end of that, share three questions that they believe most people have on their minds:
Using the uncertainly of Brexit as an example, Vinay said that the beauty of a probabilistic or a stochastic approach is that you don’t need to worry about such things. He explained that the capital markets assumptions that are provided to Envizage by their client are already forward looking assumptions and these are the assumptions that the large bank or the large insurance companies that they might be working with are using to run their business on a much larger basis. he said that obviously a lot of thought has gone into those assumptions to make sure that various probabilities of various different types of Brexit have been fed into those assumptions. So, if they are using that, they are effectively putting at the disposal of the lay individual a degree of sophistication and complexity that:
He added that it’s the same sophistication that the large institutions are using to manage their own balance sheets and their own capital adequacy and solvency.
In the first phase of Envizage’s work with them, which happened over the latter part of last year into early this year, they pointed the engine at their existing customers as well as existing would-be customers that hadn’t quite made the conversion leap, had registered but hadn’t funded the account. Vinay said that the results were really quite exciting. What they found was that existing customers found this offer to be extremely helpful and a different way of thinking about their futures that exposed to them many more needs and potential pathways they could pursue, giving more opportunities for moneyfarm to satisfy into those. On the other hand, he said there would be fence sitting customers at least, a section of those felt the confidence to progress with the journey. So, from their perspective, they viewed this as hitting all the right buttons, in a world where the top part of the market of the confident high intent, what he calls the Hargreaves Lansdown type customers, are very saturated in terms of offers, whereas his clients are now able to go one level down and start serving a less confident, less decisive type of customer.
For more information about Vinay’s talk, visit envizage.me
Julie believes that we the need to approach our careers in a very different way. She said that if you are entering the workforce today, you would need to plan for a 70 year career, not working until the age of 70, like our parents did, but for 70 years. This means that if you’re in your 30s and 40s even, we’d have longer working lives ahead of us than we’ve even lived. She said that the World Economic Forum estimates that the percentage of jobs today’s children will do that haven’t even been invented yet will be 65 percent. So the brand new is becoming the new normal at work. Plus, people are only averaging 4 to 5 years now in a job, we’re less patient, Millennials want to see tangible results really quickly on what they consider their career investment, and Julie thinks they’ve got a whole different vocabulary around rewards. They prioritise meaning, impact and experience. She therefore doesn’t think that we’ll have just one career anymore but maybe 10 to 15 ‘careerlets’, which may be at different firms, or might be roles within the same firm.
Julie that the way we use technology in the workplace is kind of archaic and that we need to think about how AI and technologies can solve day to day career and work challenges. She said that one of the biggest untapped areas of know-how is all of your colleagues, especially the ones you don’t know yet or you’re a little bit afraid to approach, and so there are barriers that really stop advice flowing in companies. The first thing is time and then how do we know who to go to when we need advice? Also, some of us lack confidence, especially if you look different or sound different, as it’s not always easy to speak up. In fact, Rungway research found that half of us have something that we want to tell a manager but we don’t feel we can. Julie therefore built Rungway, where people post questions with the option of anonymity and then they’re matched with colleagues to help. The result is that companies see really interesting dynamics, especially around gender. For example, when women make up a third of your workforce, they ask two thirds of Rungway questions and the things that they ask really reveals important unmet needs. Therefore, Julie believes that companies have to embrace these new ways of thinking and helping and actually create an environment where people can genuinely speak out.
Julie thinks that the nature of networks and teams is going to change beyond recognition on the back of AI and that we’re going to need to learn from each other in ways that we haven’t had to before. Therefore, AI will be a really important partner. However, she believes that our human skills of critical thinking judgement, emotional intelligence and frankly just managing people, will still be needed so that we can challenge the outputs of some of these models, in the same way that you wouldn’t follow a colleague blindly. She added that we’ve got to inspect the outcomes of what AI is is telling us, we’ve got to enlist more diverse people to inspect those models and we’ve got to guard against bias. Some of these challenges are not a technical problem for quants, they are a business and leadership issue. Therefore, we’ve got to hold ourselves accountable to implement and innovate AI properly. Julie therefore sees AI as a new colleague that needs a bit of feedback and guidance. Blind reliance on big data is not the answer. We’ve got to understand the personal and the individual. Yes AI will change the world of work forever, but it’s a new colleague, which can become a good friend in how we ask questions and support each other to learn challenge and thrive.
For more information on Rungway, visit rungway.com.
Yasin explained that Arabesque started at Barclays in 2010 before a management buyout in 2013. He said that the company is built around sustainability – for which they have Arabesque S-Ray, which Yasin described as one of the world’s leading providers of sustainability data on companies. They look at over 7000 companies on a daily basis and give a score between zero and 100 on many different sustainability metrics. This is then used in the asset management business of Arabesque to first understand the sustainability of companies and build investable universes based on these sustainability metrics, and then, once they have eliminated or removed and identified those poor performing companies on the sustainability issues, they then apply quantitative methods.
Yasin was at the form to talk about a new fund that Arabesque were launching, which is based on AI technology, which he said they had been working on for several years adding that the purpose of the engine is to come up with investment recommendations using, what technically they refer to as massively distributed large scale computational graphs. Yasin explained that essentially this means that if you imagine a massive network of many different data sources connected to the many different machine learning algorithms, i.e., economic models and mathematical models, all with the purpose of understanding the underlying drivers within the financial markets, which then could be coming up with investment recommendations for a given stock on a specific day usually a probabilistic weighted investment decision recommendation.
To explain their reasoning behind developing the AI fund, Yasin said that we’ve seen a couple of developments in AI in the recent years, which we didn’t have before. One is the increasing amounts of data that we can access can actually process, and the other being the computational performance and technology of the underlying technologies able to process this big data. Furthermore, there is the advancements in recent years in artificial intelligence research, principally narrow AI, specifically focused on machine learning algorithms, cyber computational statistical machine learning algorithms or some kind of multilayered weighting systems like neural nets etc. Yasin said that these have been around for a long time but we’ve never had the kind of combination of technology, data and now advancements in these underlying algorithms. He sees most of the market shifting to AI driven funds in the future, except for maybe illiquid bespoke type of investment decisions or products, but anything on the liquid side, he sees over the next 5 to 10 years move into artificial intelligence based decision making and in fact, at the CFA UK Forum, he said there were plenty of examples, not just on investment decision making but on a holistic side or on a company wide side, i.e., a lot of way that people have used to being able to do their work, process and act and make informed decisions, you are now seeing a lot of artificial intelligence and automation help with more informed decisions by bringing different data from different part of the company to make more informed decisions.
With all this in mind, Yasin is actually most exciting about what Arabesque are developing for the future and where they want to take their AI engine over the next two to five years. They have over 50 R&D projects in the pipeline covering many different areas. He describes what they have developed as a brain, whcih they have the first version of, and a massive network, which they are constantly adding to, and therefore it’s constantly evolving, with more data sources, different learning algorithms and more economic models. He said that everyone at the moment is very much on narrow AI, which is the space that Arabesque is in as well, but that is focusing on a very constrained problem or very specific question. The artificial intelligence system can only do really one thing but general artificial intelligence is the goal that they have, although this maybe 20 years down the pipeline before they actually see something which is general artificial intelligence.
To find out more information on this topic and the work Arabesque are doing in this space, visit www.arabesque.com
Clare joined us after just finishing a breakout session on the topic of ‘Using technology to prove and improve investment skill’.
She started by explaining that Essentia Analytics does behavioural analytics for fund managers, aimed at helping a human make a measurably better investment decision, which for fund managers, she said that making good investment decisions is of course their entire job. They use machine learning to analyse all of the fund manager’s historical decisions and identify behavioural patterns that in some cases are very good and evidence of strong skill, and in other cases are destructive and are things that they would probably fix if they knew about them. She described it as acting as a digital mirror to a fund manager and then using technology to help the fund manager mitigate these biases that he or she is showing.
Clare said that AI can mean a lot of things and so she began her presentation at the forum by talking about the difference between augmented intelligence, which is what Essentia does, which she explained is computers helping humans do a better job of being humans, as opposed to autonomous intelligence where the computer makes all the decisions, basically doing the job for the human and freeing the human up to do something else. She said that the investment management industry is very interested in this topic but is not very sure how to apply it in day to day life.
Clare said that Essentia’s application of augmented intelligence really goes back to the first principles of any fund manager, i.e., an understanding of what is it that I am doing that is working and what is it that I am doing that is not working and in what circumstances. Whilst that’s a complicated question to answer, Clare said that AI makes it possible.
Clare talked about PWC study from a couple of years ago that’s compared a number of different industries and the extent to which they rely on human judgement versus machine algorithms, and it showed that asset management was far and away the most reliant on human judgement. However, Clare said that humans are biased, particularly when it comes to making financial decisions, as behavioural finance literature will show, and professional investors are no exception – if they are human then they have biases. Therefore, she feels that with an industry that is massively reliant on humans for judgement calls, that is made up of biased humans, it means that there’s an opportunity for AI to help fix it.
Clare also talked about behavioural patterns and to explain it she said that Essentia analyse the data of every trade that a given investor has ever made and what you find is, as you as you would if you analysed every golf swing that a golfer has ever made, or every stroke of a tennis racket, there are patterns to how we do these things – every human has patterns – and sometimes those patterns are profitable and they would be something you’d want to ‘brag’ about and sometimes they are not. She said that there are some very common value destructive patterns that they see, like holding onto losers for too long – something she said is well-documented in behavioural finance. Everybody knows that they have a tendency to do it but they haven’t actually seen themselves in the mirror doing it and even if they have, they then don’t necessarily have the wherewithal to stop themselves from doing it the next time! So, what Essentia do is identify the pattern in the first place and the circumstances in which it’s most likely to occur and then when they see that pattern arising again, they ping the fund manager what they call a nudge, but basically just an email, saying “heads up, here are some names in your portfolio where that pattern is starting to show up again, not telling you what to do. However you might want to answer these three questions that you told us you wanted to be asked in these circumstances.”
On average, Clare said that they have uncovered for their clients over 90 basis points of Alpha that they’re giving up to either holding onto losers too long or not increasing position sizes when an historical pattern is shown that actually that’s their buying opportunity.
There are free white papers on www.essentia-analytics.com and blog post for further reading on applied behavioural finance and continuous improvement and performance.
Tim said that like a lot of the emerging technologies in the space, AI, Cloud etc., Blockchain suffers a little bit from a hype cycle that gets people very excited about it and often it’s not able to deliver and then people get a little bit disillusioned with it. He therefore wanted to focus his session on how we can get things into real production i.e. how could an asset manager or a custodian or a prime broker or a fund administrator, all the big players in asset management business, how can they genuinely jump on and generate business value.
To give an example, Tim focused on collapsing the really significant reconciliation burden that we see in the financial services industry and particularly in the asset management industry, so be very specific with big hedge funds, big asset managers have multiple service providers in the form of custodians and fund administrators and prime brokers as examples and auditors as well and right now they connect bilaterally, so there’s lots of slow asynchronous expensive difficult to maintain connections that send data. All those need to be reconciled and check made to know if you are holding the same piece of information. However, Tim explained that with blockchain, you can collapse that burden instead of posting the information to multiple providers and just post it once and then give them access to that, although of course it has to be secure and validated. He added that there needs to be an audit trail and that’s what blockchain gives us, it gives us that construct to allow us to collapse that burden and the outcome is that everybody knows much much sooner where everything is and what’s going on in their world. Therefore it provides a real benefit both in terms of efficiency, cutting costs and in not spending time reconciling but also deployment of capital. He said that the lifeblood of the asset management industry is investing and if they have a better handle on their investment and capital position, then they can have a better handle on how to invest.
Tim’s hopes for the future are that more players are brought to the table, where they’re all sharing data in this way. He believes that would lead to some really meaningful outcomes, like the notion of a distributed investment book of records – a held notion of the entire state of the asset management industry that regulators have access to and that is secure.
For more information, visit DrumG.com
Geoff shared a number of usage examples from his talk of typical examples of AI applications in asset management, insurance, retail banking and corporate and wholesale banking including:
Geoff felt that AI has been looked at a lot in the past, with a lot of examples of using AI in the labs and proof of concept but now it’s really coming of its own and as he showed, there are a lot of implementations of AI out. He said that one of the questions asked during his presentation was ‘Why now?’ and one of the reasons is because of the availability of the computing power today that actually handle a lot of datasets.
To get a copy of Geoff’s presentation of to answer any individual questions, visit www.htfcorporate.com and get in touch with him there.
Finally, if you subscribe to the show, please can you give it a positive rating and review on iTunes in particular as this helps it up the charts!