Derivatives traders and technology developers are only scratching the surface of where artificial intelligence and machine learning can take them, speakers said at a recent Futures Industry Association conference panel.
The problem, they said, isn't the amount of data that can be used in AI or machine learning; it's weeding through all the data to get to what traders need to know.
"There's a mountain of data," said Scott Hesse, lead data scientist at text analytics provider Amenity Analytics, New York, during the FIA panel in Chicago on Oct. 18. "You need to understand what the trader wants. You can't just throw paint against the wall and see what sticks. You need to get up to speed on trading data and make sure it can be modeled with precision testing, and then can implement the results. Machine learning is more than just an automated tool; it's where can it help."
Added William Dague, senior data scientist at Nasdaq Inc.'s machine intelligence innovation laboratory, New York: "We're at a point where we're still trying to figure out the language, getting the right data, getting it clean, having alternative content."
Mr. Dague said while alternative data — information not normally considered in finance — could ultimately be used to find hidden alpha in an investment, investors and traders need to know how to apply that data.
"You can train an algorithm to detail the kinds of cars that are parked at a Walmart," Mr. Dague said. "You can have algorithms that analyze voice and social media. But it has to be able to help someone understand the facts on the ground beyond just pricing, ways that data can help find that edge, that alpha."
"There's nothing worse for a trading firm than bad data," said Rob Creamer, president and CEO at Geneva Trading, Chicago, and the panel moderator. "We're still largely focused on good, clean market data. Now you're adding more layers, knowing what sources they're coming from."
However, the next 10 years will see many of those issues resolved as more developers and traders become better at understanding how to use the technology, said Cris Doloc, CEO at ALGOMEX LLC, a Naperville, Ill.-based firm that applies AI to precision medicine. "The next decade will be the decade of data," Mr. Doloc said. "What you have, what you need, how you use it will all be clearer."
Similar to the early days of the internet when its application to a variety of tasks was not yet known, the application of AI and machine learning to derivatives is only being explored — more so by vendors than by the industry in general, Mr. Doloc said.
"The industry is far behind the vendors in what the vendors can offer and what the industry needs," Mr. Doloc said.
Added Nader Shwayhat, CEO at Green Key Technologies, a Chicago-based developer of voice-recognition technology. "The vendors are so far ahead of the industry, but 10 to 15 years from now, AI will be the norm."
In derivatives trading, the speakers said, there will always be a human element regardless of how much trading activity is taken on by algorithms and computers because people ultimately will be responsible for the actions taken in AI.
"Vendors that are creating ways to replace humans in trading decisions are trying to solve a problem that doesn't exist," said Luca Lin, partner, Domeyard LP, a Boston-based hedge fund focusing on high-frequency trading.
The future trading station, said Mr. Shwayhat, is "a human who's working with four, five, six machine-assisted data distributors. But there will still have to be a human making the decisions."
But Mr. Doloc said that future AI applications will be out of the control of traders and vendors. That's because out of the three steps of AI development — automation, innovation and discovery — the third step is the one where people don't control the outcome.
"We're still struggling with the automation ... and what to use the automation for," Mr. Doloc said. "The discovery is finding new uses for automation beyond that, and with AI, that we may not control."