Machine learning is everywhere these days, and finance is no different. Both large banking institutions and scrappy startups are attempting to “disrupt” money with artificial intelligence (AI), collectively helping make AI an integral part of the industry’s future.
As the head of the commerce platform machine learning team at Square Inc., Marsal Gavalda helps lead the technical aspects of the Jack Dorsey-founded payments platform. Gavalda and his team apply machine learning and automation to Square’s payment products, which in recent years have aimed to revolutionize the way merchants conduct transactions.
SEE ALSO: How a New AI Breakthrough Could Undermine the Financial Industry
Observer spoke with Gavalda at this year’s AI in Finance Summit by Re.Work, where he described the current climate at the intersection of tech and finance.
How did you start your work in AI and finance?
It’s actually an interesting story. I got into AI because I got interested in language technologies. And I got interested in language technologies, I think, because I grew up in Barcelona—Catalonia, Spain, where we speak Catalan as opposed to Spanish. As a child, I got very interested in languages but also in computer science. One day, my dad showed up with a HP-85, one of the early home computers, and that’s how I got into it. I did computer science as an undergrad, but then I realized that there was this mixed in field called computational linguistics, now better known as language technologies, which is computers trained to do things like speech recognition, machine translation. And it was really interesting because 20 years ago, you would do that almost sort of by hand, writing these rules, that say “well, an English sentence is made out of a noun phrase and a verb phrase.” But very quickly this was taken over by a much more bottom-up, data-driven statistical approach, which is what is what machine learning is about. So machine learning is learning directly from data. And this is now how you can do speech recognition and very high accurate machine translation. So, that’s the field I was in. And then came the opportunity with Square to apply machine learning to financial transactions, and also to analyze the interactions that sellers and buyers have through our platform.
Being at Square at the moment, in particular the commerce department, can you tell us what that entails?
At Square, the overarching purpose is economic empowerment. And therefore, anything we do, we do with that sort of objective. So with machine learning and artificial intelligence, it’s no different. We are trying to find the best applications of this new technology in a way that empowers our sellers and our buyers to become part of a financial network and to thrive in this financial network. There’s many great examples. There’s Square Capital, which is the arm within Square that uncovers opportunity within our sellers and issues loans so that they can, for example, grow their business.
And so what does that look like, day-to-day, to work on?
Well day-to-day, it’s really a matter of prioritizing. We do know that the data that we have is extremely valuable to uncover these opportunities and provide a better experience for our sellers and our buyers. It’s almost a matter of trying to see from all this data and these models that we develop: how can we surface that as a product feature to our customers in a way that is useful and relevant to them?
A lot of your customers being merchants themselves, how valuable is this data to them?
Sellers are very interested in how their business is going, and also how it might be improved. We have features, for example, that are becoming more personalized. Recently, we launched the ability for when you go to your items page, you can see suggestions about what other items you may want to add to your catalogue. This is a way to make the checkout faster because you don’t have to enter a price, you can just sort of tap on an item. It also allows us to give more detailed analytics and reports and trends to our sellers.
Within finance itself, it’s a broad field with historic institutes. How would you say AI is transforming a very regulated and closely watched industry?
Yes, there are some constraints, some of them for good reason. For example, we certainly want to make sure that the models that we develop are compliant with any new regulations such as GDPR in the European Union or SECA in California. We want to make sure that the models that we develop don’t have a so-called disparate impact on protected classes. And so, we have compliance revision that also makes sure of that. More generally, we are certainly aware of the huge transformation that AI and machine learning is having not only in the finance industry but in society at large. And for that we certainly want to be part of it and help shape it in a way that we’re going to be happy with, in the sense that we want to be building a society that we would all want to be part of and not some sort of strange dystopia where we’re being ruled by these models.
And what do you see as potential concerns in applying machine learning to finance, where there are so many security concerns about sensitive personal information?
Certainly aspects around data privacy is something that we take very seriously. In the entire history of Square, I’m not aware of us ever sharing our internal data with any third parties. So if you have a transaction within Square, we’re not going to monetize that. Beyond that, as I mentioned, is trying to make sure that there’s no such a thing as this disparate impact on protected classes, but also that we build models that we feel comfortable with. And we always have an eye out for the mistakes that it can make, be it a false positive or a false negative. We always look at the consequences of that. And before we release a product, we make sure that we have a sort of graceful way of handling that, specifically any of these models that become visible to the consumer. We always have some design and some way of basically lowering the impact of a potentially incorrect decision.
Square has a fairly consumer-facing presence, especially via the Cash App. For people reading this interview in five or 10 years’ time, what would you predict would be their thought on the way Square transformed financial services?
What we hope to achieve with Square is this purpose of economic empowerment at many different levels. As a seller, you want to grow your business. As a buyer, you want to have more options and more rewards. So something that we’re doing with the Cash App, which is unprecedented, is having these rewards attached to a card that has no maintenance fees and is completely free. So, I think that what we would want to be remembered for is this company that really tried to create a more honest, transparent financial network that everybody can be a part of.
And this includes building an ecosystem around products, like the recent automated Photo Studio service that Square launched for sellers?
Yes. We always try to sort of make it easier for our sellers to improve their presentation. In this case, it’s a service where we generate professional-looking images of your products. Basically, there will almost be this dichotomy between brick-and-mortar and online. It’s going to be a much more fluid sort of way of how buyers and sellers interface and how commerce occurs.
So what’s next for you that you can tell us about?
There’s a lot of interesting stuff in the pipeline. I can not disclose it now, but I’ll give you a hint which is—this is a public fact—a recent acquisition of Eloquent Labs, which is a spin-off of Stanford devoted to cutting-edge conversational AI. We do think that that’s an area where we can come up with some really interesting products.
Outside of the fintech space, as an AI expert, what are you most excited by at the moment?
There are many areas. I would say, probably augmented reality, virtual reality is a very interesting space. And my longtime passion of languages in the area, of not just speech recognition and some natural language processing, but true natural language understanding. This understanding of not just correlation, but also causation, and a deeper ability to have conversations and feel that this system really understands what’s happening.
Credit: Google News