In a recent blogpost, Bart Joris, head of FX sell-side trading at Refinitiv, noted that currency traders perform better when artificial intelligence (AI) techniques are coupled with human insight. He added that new skillsets were becoming more important, to enable these individuals to use AI and machine-learning tools to make the best decisions.
It is a topic that is exercising many in the FX world right now. Artificial general intelligence (AGI) – a type of AI that has the ability to understand, learn and apply knowledge across a wide range of tasks, in contrast to narrow AI, which is designed to perform a specific task – could theoretically engage in conversations with traders to help then fine-tune their decisions, suggests Thomas Friesleben, managing director at StoneX Pro.
“This could involve the AI understanding the trader’s goals, strategies and risk tolerance and then providing personalized advice based on its market knowledge,” he explains. “AGI could also automate order execution and currency monitoring, although some forms of these tasks are already being performed by narrow AI.”
AI tools can narrow the gap between those traders with access to Bloomberg terminals and those without, by compiling and sorting data quickly and easily, says Allen Li, global head of e-risk for FX options at HSBC.
“Access to data will be among the first areas to benefit from AI,” he says. “We will be able to access and interact with data digitally in the most natural way – natural language.”
Using an AI hierarchy will create conversations that enable decisionmakers to better focus on specific or unique data and cut through the noise to develop solutions that improve transparency, says Hans Jacob Feder, managing director and global head of FX services at MUFG Investor Services.
“AI will likely augment existing data sources, including Bloomberg terminals, and create new opportunities for analysing and processing data from other types of sources – all of which will help traders,” he adds.
Race to full integration
There seems to be a race on to create fully integrated software that allows a trader to input systematic rules that can then be backtested, optimized and automated, according to Chris Weston, head of research at Pepperstone.
“For the large majority of traders – who aren’t engineers or developers and are not versed in asking exactly the right question – ChatGPT is pretty painful and requires numerous tweaks to get anywhere near the right code to use for the platform,” he says. “Then there is a good chance you will need to use logic to see where the error lies.
“This is a huge step forward, but it is not the panacea that makes automated trading easily accessible to all.”
However, order management system vendors and trading platforms are starting to make these capabilities native in their products. With improvements in natural-language processing, it is expected that this type of interaction will become more widespread. That makes it possible that traders will no longer process individual orders, but will rely on macro instructions in natural language to build their portfolios, automate order execution and manage their risk factors.
Roel Oomen, global head of FIC quantitative trading at Deutsche Bank, says the bank is focused on AI’s ability to summarize information in a user-friendly and bespoke manner, enabling sales and trading teams to track live market sentiment, generate market colour in a targeted manner, and combine real-time newsflow with verified research and central-bank statements to improve risk management.
“This is also about utilizing the teams’ specialist knowledge to filter out inaccurate AI model predictions and provide a feedback loop to improve model training,” he adds. “The human feedback loop is critical for reliable AI development.”
The flipside
There is a flipside to the progress, and those operating in the industry seem well aware of it. Over-reliance on technology in FX trading may lead to errors in decision-making or to individuals overlooking better options, cautions Kate Leaman, chief market analyst at AvaTrade.
There are also concerns that future AI tools may offer up investment advice, although the legal and regulatory ramifications of this remain unclear, notes David Morrison, senior market analyst at Trade Nation.
“ChatGPT is quite adamant that it doesn’t have the ability to predict stock-market movements and doesn’t have real-time data, but there is no reason why other generative AI units cannot be plugged into real-time market data,” he says.
Morrison adds that it is crucial to consider the potential downsides of relying too heavily on technology in a trading environment. He recalls that in the high-frequency trading market, early concerns were substantiated by ‘flash crashes’ that caused panic among investors.
“Over-reliance on technology can also introduce risks such as algorithmic errors, system malfunctions and unforeseen market conditions that may not be adequately accounted for by AI models,” he adds.
In addition, systems can fail – a technical glitch could lead to substantial financial loss in a high speed, high stakes environment such as FX trading. And a further concern is the risk of homogenization.
“If everyone is using similar algorithms, it could lead to markets moving in lockstep, increasing systemic risk,” concludes Friesleben at StoneX Pro. “Over-reliance on technology could also lead to a lack of human oversight, which can be critical in identifying and correcting errors the system may need to recognize.”
