The Autumn London ThinkTank was hosted at The Curtain Club on 25 October 2018. Tony Guida, senior investment manager in quantitative equity at the investment arm of a major UK pension fund in London, chaired the event.
We can all agree that machine learning and AI are growing massively, with many asset managers incorporating the techniques into their investment strategies. machineByte invited several ML/AI thought leaders from the UK to participate in our first ThinkTank meeting, where they discussed key topics of this moment.
Finding the right research materials
Machine learning and AI are unique tools and their adoption is now global, but practitioners are still unsure of what everyone else is doing? What are the new initiatives?
One of the attendees illustrated that point, noting the challenges that exist in the space: “Before there is a big burst, we need to make sure we have a lot of real and educational materials.” Within the tech world, companies such as Google, Uber and Tesla are all willing to share their research materials and papers, but in the financial industry the attitude towards sharing research is more closed off. That attitude was one of the reasons why ThinkTank Chair Tony Guida started writing his recently released ML/AI book.
Machine learning and AI are technologies that reflect the world that we’re living in, but the finance industry still needs to catch up. There are a lot of companies that claim their portfolios are AI-driven, but often they are using stock selection algorithms rather than full AI strategies.
This means the market needs to be looked at with some scepticism. More machine-learning presentations are given by investor relations teams and not by the actual tech teams. According to one allocator, that the biggest risk from an allocation perspective.
Finding the right talent
Banks and asset managers are hiring data scientist and AI engineers from Silicon Valley, hoping that pedigree will bring new ideas to the financial services world. Often, these new hires are brought in as MDs, but do they actually know the financial market well enough? One attendee used an example of a potentially great new hire, who worked on a “new” project that was basically invented 25 years ago.
So how do you find the right new talent when AI engineers know AI, but don’t know how to apply the technology correctly in the financial space?
Chinese applicants hired more often, according to one panelist, because of their massive advantages in compiling data. China doesn’t have many privacy laws as compared to Europe or the US. A Chinese insurance company can analyse all the information about a new applicant without any limitations, which gives the engineers that work there an advantage.
The AI hype
AI is cyclical. One attendee mentioned that when he started his Ph.D., AI was on a high note. During the dot-come boom, AI experienced a lot of unwarranted hype. “There were some real applications that AI was actually achieving,” he said. “And so, people hyped it up too much and then it didn’t live up to expectations.”
“As an academic it’s really hard, because you find that during boom periods, there’s money thrown at you, but when it turns into the AI winter, funding usually dries up.”
During the slower periods, computer science departments at universities are typically merged into other departments due to lack of interest and funding. “So, you have these cycles and it’s precisely the type of hype we’re going through again. Let’s use AI, because it can just solve anything,” he joked.
Market data is very dense but very noisy. Though there’s an abundance of data available, often this data is not usable.
Financial markets also suffer from limitations in data collection: there’s no way to run experiments how different events affect financial markets because market data are linear. In other scientific fields, like medicine, practitioners can run several scenarios to gather more data, which makes their models stronger.
One attendee mentioned that he is replacing large blocks of code with something that can be validated, back tested and which practitioners can fine-tune. “You bring in validation, you bring in a sample, you test it, you ensured all of these steps happen and you can compare it very systematically in parallel on the same data,” he continued. “And it just brings a great methodology to the process of writing algorithms. You’re far more systematical and that’s how you improve.”
No commitment 30 day free trial
We deliver education to quantitative professionals on the latest innovations and developments in machine learning in investment management and other industries, such as tech, healthcare, transportation and education.
Get the latest exclusive news, innovations and approaches at financial institutions, tech firms and companies across the whole spectrum of machine learning.