Conference Dailys TRADETech Daily 2019 | Page 21

THETRADETECH DA I LY in-depth THE OFFICIAL NEWSPAPER OF TRADETECH 2019 “One challenge to keep in mind is a trade off in the realm of data management. There is a fine line between cost savings on the employee side and incurred costs of large data mining systems.” LARISA MATHUR, VNR PARTNERS AI is not going to be able to bail us out when such instances transpire and if everyone uses AI then who has the advantage? There is no consensus on the best AI poker programme, and it’s not clear which one would win in a “head to head”. Tom Doris, chief data scientist at Liquid- net, points to a consensus on the buy-side that AI is the “way of the future” and that the industry will see a fundamental trans- formation in the next three-to-five years. He notes, however, that fundamental trans- formations often take a little longer than expected and suggests seven years as a more realistic timeframe. Blind spots, of course, remain. AI offers the possibility of gathering online user feed- back for new product launches to quickly determine whether the company behind it is looking at a hit or a miss. AI has a particularly hard time spotting sarcasm: what reads as a straight thumbs up from a widely followed social media pundit may in fact be a wither- ing thumbs down. Problems such as sarcasm, Doris says, mean that “a high degree of scepticism is still needed. The human decision-making process is still central to wealth and risk manage- ment. I wouldn’t want my money managed by algorithms.” AI, he says, is nowhere near achieving a human level of understanding. The hypotheses and assumptions that are used by AI, he says, will continue to depend on what is important to human beings; the in- dividual and collective mechanisms through which we as humans use to adjust the narra- tives which we construct to understand the world, and the trading universe, are still far from being understood. Doris also takes a different view to Smith at Rosenberg on the impact of AI and automa- tion for buy-side jobs, arguing that a lot of desk-assistant type roles have been down- sized. Consolidation, however, has been the leading driver of lower headcount. “Anything that can merge will merge,” he says. Doris suspects that this process may be close to its limits, with buy-side firms now “about as skinny as they can be.” Managing data sets Paul Spencer, chief operating officer at Velocimetrics, has been in discussion with developers of AI systems and the trading organisations who intend to use them. He consistently hears that the recognition that AI and machine learning can only be effective if the systems receive an accurate, real-time view of the transaction flows as they are occurring within a complex trading platform. “The ability to process time-stamped trade flow events captured independently from the platforms that are being monitored is of paramount impor- tance,” he says. Larisa Mathur, managing partner at VNR Partners in Miami, cautions that developing artificial intelligence projects “can get quite complex and go out of control quickly.” VNR Partners is a proprietary algorithmic forex trading shop, which has had success in using advanced methods in parts of the process where speed and dependability are critical, yet outcomes are clear. “If there is clarity of thought and re- quirement placed to AI, and a well-defined context, AI projects help support the overall strategy,” Mathur says. “However, it takes a lot of work. One challenge to keep in mind is a trade off in the realm of data management. There is a fine line between cost savings on the employee side and incurred costs of large data mining systems.” Rosenberg’s Smith agrees that time is a key constraint in determining whether a data set should be used. He is confident that AI will have a role in the future in doing this. “If we can automate it away, we’ll try to do it”, while Doris at Liquidnet believes that the debate on AI has progressed from 18 months ago and is now more nuanced than the previous positions of total acceptance versus outright scepticism: “You never progress from a posi- tion of fear.” Issue 1 TheTradeNews.com 21