Antonio Simeone, co-founder of Euklid, which manages savings and investments through thousands of algorithms, tells Arab News how his company is developing code that aims to understand not only the vagaries of the market but also the psychology of the trader. Can you describe the likely application of your idea? That is to say, are you offering the technology to existing fund managers or other investment companies, or do you plan to perform that role yourself — deploying capital for clients in investments picked by algorithms? We have already established a fund in Luxembourg and recently started raising funds from both qualified and institutional investors. The AUM (assets under management) in our fund are then invested utilizing our artificial intelligence (AI). Our AI studies the psychology of traders and recognizes the fingerprints that they leave behind, besides also being able to identify patterns and micro-patterns which can’t be perceived by the human eye. Due to our entrepreneurial spirit, we chose to set up our own structure and manage money directly instead of licensing our technology. Part of a fund manager’s job is to predict investments that will perform well in the future, not just collate those which have done well in the past. While algorithms can help with the latter, are they as effective in looking ahead? By using our algorithms, we aim to understand the traders’ psychology regarding a specific stock. Our experience has shown us that each trader leaves some kind of traces and that these signs are written in the historical series. We aren’t managers but scientists. What we do is completely guided by artificial intelligence, thus removing human discretion in the investment process. Particularly, our AI is capable of recognizing stock fingerprints that traders leave behind and, by analyzing these and a number of other variables, predicting the changes in value of each stock in our portfolio. We have developed around 45 algorithms that have been customized for every single asset that we trade. Our portfolio is made of 184 equities among the most liquid on the markets, 70 percent being US equities. The algorithms are based on biocomputing, a science linked to maths, physics and biology. In order to keep the learning process efficient, an optimization process is ensured by swarm intelligence, neural networks and genetic algorithms. We could say that we create the basic foundation and then it is up to AI to create other structures autonomously and independently. Let me simplify this concept for you: It is as if we had thousands of traders at our disposal, who are, however, virtual. The strongest or the best are those that stay alive, and those who are not simply die. Do you have targets in terms of projected assets under management? Our fund launched in mid-August and therefore we don’t have a vast AUM at the moment. However, many qualified and institutional investors who have been following us for a long time are really interested in our activity and we have received many soft and hard commitments. Our objective is … to reach the maximum amount (around $13 billion) manageable with our algorithms in three to five years. This limit is due to the fact that the algorithms trade exclusively blue chips and highly liquid stocks. However, the AI is created to understand and predict the market’s psychology, and it would start to have issues when the AI itself starts influencing markets. How far away are we from the point at which algorithms replace fund managers in the same way that other functions in different sectors have been made redundant by technology? This is the reason why we do not have any management fees. We are a team made up of just a few people but our technology acts like thousands of “virtual” traders who work 24/7 for us. They are able to observe an asset from many different points of view. We are the first fund to use both AI and blockchain. But, honestly, the financial industry is rapidly evolving and getting more and more software-based; Thus this scenario is not that far from the current practice as we may wrongly think. How can algorithms make sense of the extremely volatile and illiquid markets such as in the Arabian Gulf, where there is little raw data to process? I happen to think about the cryptocurrency world. When we first started our algotrading activity on bitcoin, we had little and unreliable data. Only three years of historical series. Despite this, our algorithms were able to understand the trades’ psychology in a very accurate manner. But that was a very volatile and predictable scenario, and I don’t think we could achieve similar results with other assets. The traders’ psychology was redundant and very simple; even the alleged manipulations were easily predictable. But, right after the bubble burst — or maybe right after the features were issued — the market sentiment really changed. Algorithms keep learning but this world seems less attractive and it is impossible for our algorithms to operate with more than $50 million because of the limited market capitalization. As far as the Arab market is concerned, we are still studying it but we have many problems in searching and computing data.