The finance industry is just at the beginning of an unprecedented era of disruption. The accelerating speed of technological change and the potential for automation promises massive changes for banks, investment firms and other parts of the financial industry.
Much of the discussion about artificial intelligence and machine learning has focused on manufacturing. Technological innovation manifests itself most visibly in automated processes that until now were the domain of manual labor.
But it is only the first step. The next wave of intelligent machines is threatening to disrupt the white-collar worker and the more creative thinkers.
In the staff-intensive banking industry, for instance, humans are the main cost center, and the ability of blockchain technology to replace some of the workforce is one of the top initiatives pursued by senior executives in shaping their business.
Impact on hedge funds
Even the world of alpha creators in the alternative investment space may not be immune from the rise of the machines, speakers at the recent Cayman Alternative Investment Summit noted.
Raoul Pal, Cayman-based CEO and founder of the Global Macro Investor and Real Vision, says the best hedge funds have a process, and that means that the best hedge funds can be replaced by a machine.
He points to algorithmic trading and high frequency trading as the first highly successful examples.
“I think most trading will be replaced by machines, but you need to understand that what that means is the hedge fund manager of old just gets replaced by a programmer. So the hedge fund manager will need a different skill set to focus on systems-generated returns.”
While Pal believes computers are very good at processing current information, he thinks they are “terrible at looking into the future,” a time horizon much more suited to human intuition.
“As we see the investment management industry shift with the flow of pension money coming in and becoming more short-term in nature, humans are going to do what humans do well: have a strategic vision for the world,” he said in a panel on how technology affects the alternative investment industry.
Where machines fail
A recent example for the current limitations of machine learning are Brexit and the election of U.S. President Donald Trump, which machines were unable to predict.
“The simple reality is machines are not in touch with human emotions and what people feel and think,” said Suryanshu Mishra, the head of Hedge Fund Administration at Deutsche Bank.
Even with the best polling mechanisms, it would be extremely difficult to gather accurate, reliable information from 7.5 billion people. To then combine data collection, information management and statistical analysis to make assumptions about the disruption of economies and political systems or to identify the best trading activities would require a huge amount of human intelligence and research and development.
Simply basing a trading strategy on predictive analysis with limited and partially unreliable data sources is a flawed approach, Mishra said.
“The investment time horizon might be two to five years, but the information time horizons that you can rely on do not extend for more than two or three days. We don’t know what’s going to play on CNN tomorrow, so forget about a year from now.”
In the debate about the disruptive force of artificial intelligence, the term is often used too loosely, said Bettina Warburg, founder and managing partner of Animal Ventures, a tech consulting firm. The strict definition means that machines would develop a consciousness and start to think for themselves.
“What we have today is much more like intelligence augmentation.”
It is not so much that things are new, Warburg said, “but we are advancing in so many different fields that gives people a sense of unease about the future.”
Smart algorithms, machine learning tools and parallel processing are great advancements that in combination will allow augmenting different kinds of human intelligence, she said, adding that rather than conscious intelligence, the aim is “artificial smartness.”
The future will therefore lie in a human-machine symbiosis that will require humans to adapt.
The talk about artificial intelligence, however, is not just speculation about the future.
Most current applications in financial services are based on supervised learning and require the interaction of humans and machines. For example, spam detection systems are trained by giving them a designated outcome: some emails that are marked as spam and some that are not. The systems then analyze a range of features that in combination become indicators of the correct category.
In the payment industry, the prevention of fraudulent transactions using stolen cards relies on a similar approach.
“We have been using this in the fraud space a lot,” said Vinay Rao, head of Risk at Stripe Inc.
Based on the large number of charge-backs in the industry, there are sufficient examples to train a machine to detect the features most indicative of fraud. “Historically these features have been human coded,” said Rao. Humans know the features of a transaction, but as frauds have evolved, humans are no longer good at combining features and determining correlations.
“That’s when you start using feature discovery methods like clustering. And then again you bring humans to determine do these clusters represent fraudulent behavior and use that in your method,” he explained.
In the alternative investment industry, the majority of back office functions are outsourced, but consultants and technology providers are looking at blockchain and artificial intelligence for automation and to answer the question of how to cut cost and risk out of the business.
Richard Scott-Hopkins, director at KPMG, said all aspects from front to back office have come under scrutiny for potential automation.
On the trading side, he noted, there is already a blockchain solution for derivatives which eliminates the intermediary and saves time by using smart contracts that settle in real time and manage margin collateral.
Even compliance officers need to become technology professionals, he said, and take the rules they have been applying for years and put them into code so that systems can monitor the trading portfolio and allocations.
And at the service provider level, CEOs need to talk to their administrators and custodians about what they are doing about technology.
In this context, asset managers right now are focusing on the challenges around the interoperability from front to back office and the interoperability between this ecosystem and the counterparties, like prime brokers, exchanges and fund administrators, added Mishra. Since systems and tools are often designed with different programming languages, it makes it hard for one system to speak to another and to reconcile data.
“Data becomes a monster unless you figure out how to manage,” he warned. “You may have the best data stored in your database but if you are not able to speak in the same language with the building blocks around you and within you, that is a problem.”
Many new solutions rely on blockchain for automation. Blockchain, the technology that is underlying cryptocurrencies like bitcoin, is essentially a distributed database that maintains a growing list of records or blocks. Although it is much talked about, the real meaning and impact often remains elusive.
Warburg said there are three ways in which blockchain is going to cause dramatic change.
First, we are moving away from an internet of information to an internet of value, she said. “We are actually able now to uniquely create assets out of digital and physical goods and transfer that value without the need for an intermediary.”
The technology itself is used to lower the uncertainty involved in these value transfers.
Second, there are many kinds of blockchain in development, with some focusing on privacy while others are more scalable or have different governance structures. “So we’ll actually see a world of thousands of blockchains that will interact and connect in interesting ways.”
And finally, blockchain will bring about new kinds of assets not limited by making what we already know better, she said, but rather “inventing a new logic to how we do business.”
Blockchain will not only lead to the creation of new products, but also new markets, whether it is around cryptocurrencies or new concepts of ownership and business organization. This could include the idea of a song as a company that doles out profit to the composer, performer, producer and other copyright holders every time it is played. Or new forms of owning an asset, like partial ownership of artwork.
Warburg also predicts a general move toward digital currencies with many kinds of currencies for different purposes. Existing virtual currencies like bitcoin and ethereum are already very different, she noted.
Deutsche Bank’s Mishra agreed, but he cautioned that there is still a long way to go.
Any buildup of a digital currency will need a lot of computing power which translates into hardware requirements and electricity consumption. Security is the other factor.
“Digital currencies will undergo a few security challenges,” he said, in terms of who controls the generation of new currency and ensuring that the mining of new tokens is not corrupted.