Cover photo: left to right: Juan Pablo Braña (CDO), Alejandra M.J.Litterio (CRO), Adrián Raguza (CEO), Nicolás Paladini (CBO). Credit: Alejandra Litterio/Eye Capital

Financial traders operate in a fast-paced, high-risk and complex environment. They deal with huge sums of money in a difficult environment and as such, are liable to make mistakes.

Many risk reduction techniques and technologies have been developed over the years but perhaps the most effective technique would be to remove the human element altogether.

This is where Eye Capital, an Argentinian startup that specialises in natural language processing, a field of artificial intelligence, comes in. The firm’s algorithms read and interpret financial news in real time, and the small team is making big waves in the world of finance.

We spoke to Alejandra Litterio, Eye Capital’s CRO and co-founder, about taking away traders’ jobs, the role of language in artificial intelligence, and the Argentinian tech ecosystem.

Origins

Seven years ago, Litterio was working on an academic project at Centro de Altos Estudios en Tecnología Informática, Universidad Abierta Interamericana to design an algorithm that could establish correlations in the markets and predict and automatically recommend whether to buy or sell equities.

The project was successful and in 2016, Juan Pablo Braña, Litterio’s husband and Eye Capital co-founder, met with two angel investors (and co-founders). The investors shared Braña and Litterio’s vision and gave it their backing.

Eye Capital officially became a company on May 4 2017, having started as an academic project six years before. Now, Litterio is chief research officer and Braña is CDO. The team, composed of the six co-founders, is multidisciplinary and it has a simple objective: to be the first fintech whose core purpose is developing AI algorithms.

Though they met at university, the wife and husband co-founders are not from the same academic background. Litterio comes from a linguistic heritage and has worked as a teacher and specialist translator. From there she moved into discourse analysis, which looks to understand how language is interpreted and understood.

Braña, on the other hand, is a mathematician and data scientist. It was his idea to design an algorithm that could read and interpret financial news in real time, which he shared with Litterio, who thought it would be interesting to ‘teach’ the algorithm linguistic techniques so it could process natural language.

What is natural language?

We do not use language in a vacuum; it is always context-dependent. The role of the linguist is to study the different components of language

Most people with a passing interest in artificial intelligence will be familiar with the term ‘natural language processing’, but rarely do we consider what ‘natural language’ actually is.

Litterio’s definition is simple: it’s the language that human beings use to communicate in everyday life. However, once we delve a little deeper and look at the core aspects of linguistics it becomes apparent that it’s more complicated than that. Linguists analyse the morphology, semantics, syntax and pragmatics of human language, and it becomes more apparent how complex that is when we try to teach that to machines.

traders

Alejandra Litterio speaking at Cloud Expo Europe in London, March 2018. Credit: Alejandra Litterio/Eye Capital

The problem, Litterio tells me, is that we do not use language in a vacuum; it is always context-dependent. The role of the linguist is to study the different components of language, how people combine them to create meaningful sentences and eventually how they use them in discourse, whether spoken or written.

And that’s important when training an algorithm. This requires ‘teaching’ it all the logical steps to identify what is a word, what type of word it is, what type of sentence it’s in and what context the sentence provides. It’s also important that the algorithm can determine if the sentence belongs to a specific field, such as medicine, law, or in this case, finance.

A multidisciplinary organisation

Of course, it’s not possible to build an algorithm without scientific and mathematical expertise. At Eye Capital, the founders embrace a multidisciplinary approach, Litterio explains. In the research and development department, the various specialists coordinate their work in order to develop new ideas, turn them into reality and record what they do from an academic point of view, justifying with theory what they do in practice.

This is the way in which more companies should work, Litterio argues. Anthropologists, historians, neurologist, philosophers and translators all have a role to play, she says, because to train an AI algorithm, you need to know how the human brain works, how the neurons interconnect, and how they operate, as well as all the practicalities of creating a working and successful AI algorithm.

The inherent risk

And it certainly needs to be successful. Financial trading requires very accurate and very cautious people and systems. There is risk inherent in the industry but limiting this is a key requirement, and this is especially true in automated trading. Risk parameters are used when training AI algorithm models, which ultimately have the potential to generate portfolios with big returns and minimal risk.

The importance of risk management is abundantly clear when looking at the proportion of code which is devoted to it. The code of one of Eye Capital’s algorithms, says Litterio, may have 3000 lines, but the logic that applies to the buying or selling of an instrument is coded between 100 and 150 lines, and the rest of the coding lines are used to control and reduce risk. Before running any algorithm with real money, they’re also tested in rigorous simulation environments in real-time sync with the markets, which helps to detect any type of failure.

But what do these algorithms actually produce? Litterio’s work is focused on designing the step by step process of how to approach the mass of text that the algorithms deal with. These have produced three tools – two financial sentiment lexicons, one in Spanish and one in English, and the Financial Market Drivers Lexicon.

Between the three of these tools, she says, the company is able to cover the analysis and interpretation of financial news in real time, in the two most widely-spoken languages in the world.

Replacing traders?

What she does believe, however, is that this general methodology – using AI and algorithms – will probably eventually replace traders

If the technology works well enough, it’s not implausible to imagine it putting a lot of traders out of work. This is by no means the intention, says Litterio. A change is needed, and any tool which can reduce risk, manage portfolios efficiently, and increase profits is a good one, she argues, but it’s not her goal.

What she does believe, however, is that this general methodology – using AI and algorithms – will probably eventually replace traders. According to Litterio, this is no bad thing – she argues that this will provide space and time for new business areas and new ecosystems to develop.

In Litterio’s long-term vision, then, there may be an abundance of fresh startups being born as financial traders are replaced by machines. But what about the here and now for Eye Capital? I ask Litterio if Argentina is a good place to be a new, fast-moving business.

There is no Silicon Valley equivalent in the country, she tells me, but during the last few years, there has been an exponential growth in the fintech ecosystem.

However, this has been mostly in spite of government support, rather than because of it. Though an Argentinian Fintech Chamber was launched in 2017, Litterio says that there is little support from the government to the sector, with most companies depending on incubators or accelerator programs from abroad. In her instance, the British Embassy in Argentina provided support and made setting up in London a relatively easy process.

Now, the company is on an upward trajectory. The next stage, Litterio says, is to break out of America and turn global. It’s already started – in 2017, they opened offices in the U.S., Spain, and London. The next step is conquering Asia. If expansion continues at this rate, financial traders might find themselves competing with machines.