Fintech’s Artificial Intelligence Revolution: The Missing Link

We stand at the threshold of an era of pervasive artificial intelligence (AI) in financial services. While it’s tempting to say, “Again?” — this time really is different.

Why? Because AI/machine intelligence will be rapidly deployed across financial institutions. And though this evolution is inevitable, we need to be mindful about the ethics of the undertaking. As an industry, we must work together to ensure careful and thoughtful AI integration.

A recent study incorporating a survey of 424 financial service executives found that “nearly seven in ten [respondents] believe AI will bring a complete or substantial change to their own jobs over the next 15 years. . . . 47% are either “Not Confident At All” or “Not Very Confident” that all material legal risks associated with new financial technologies have been properly understood.”

Where do you expect AI/machine learning technology to be introduced in your organization in the next three years?

Source: “Ghosts in the Machine,” a digital report created by Euromoney Institutional Investor Thought Leadership and commissioned by Baker McKenzie

Algorithms already steer many front- and back-office functions, as well as moment-by-moment exchange operations, including price discovery, automated trading, and order matching, among others. Humans simply cannot calculate and execute these same functions as well as machines.

Cutting-edge information technology has always been a competitive advantage in financial services. Autonomous agents executing trading and investing strategies already exist, and there is a push to turn more of the capital markets investment cycle over to machines.

The pace and magnitude of fintech investments will propel a steady stream of new and increasingly powerful AI-based applications. How well financial institutions integrate these applications will determine which of these institutions survive in the long term.

The intelligence may be artificial, but the risk is real.

Technology waits for no rules or regulations, and AI is no different. The potential profit associated with innovation economics also contains the risk that machine intelligence will be developed and deployed without thoughtful consideration of the potential perils.

And AI brings a unique set of risk challenges. If they are not well managed, we may create new and greater risks.

Indeed, it is unclear how some forms of advanced machine intelligence even work. We simply don’t understand how some machines solve problems with deep learning algorithms. Should something go wrong, we might not be able to define the problem a solution.

Already problems with technology not nearly as complex as deep learning have disrupted the markets. There have been six “stock market crashes” due largely to flawed market operations, most notably the “flash crash” of 24 August 2015. And while markets recovered from these crashes rather quickly, the immediate causes were not immediately grasped.

Another risk factor specific to AI/machine intelligence is “bias.” While it’s tempting to think that computers can be programmed to be completely objective, in fact, machine learning displays very human biases. Coupled with the well-known behavioral biases associated with investing, without proper safeguards, AI may permanently codify human biases and cognitive investing errors.

Intentions versus outcomes is another risk consideration. If we give AI autonomy, how can we be sure that the agent will carry out its activities as we intended? How effectively can we translate our intentions into computer code? That isn’t always clear.

For example, look at the distributed autonomous organization (DAO). DAO was a “smart contract” that ran atop an Ethereum blockchain. It only yielded investment voting decisions to investors who placed about $150 million into the fund. Apart from voting which investments to make, it was an entirely autonomous system. Within three weeks after fundraising closed, the DAO was hacked and the $150 million was essentially forfeit.

How did this happen? Not through clandestine “black hat” intrusion altering code. Instead, the hacker used the agent’s code in a way not anticipated by its writers. Ironically, “white hat” hackers also “hacked” the system to prevent any misappropriated money.

Many financial institutions grasp the nuanced dimensions of the risk AI/machine intelligence involves. Yet only 30% of financial executives are confident their organizations understand all the associated material legal risks.

How confident are you that all material legal risks associated with new financial technologies have been properly understood by your organization?

Source: “Ghosts in the Machine,” a digital report created by Euromoney Institutional Investor Thought Leadership and commissioned by Baker McKenzie

There’s nothing artificial about fiduciary duty.

What AI lacks is a clear, robust, and open-source agency framework to determine if and when control of an investment process should be turned over to machine intelligence.

As advisers, we have evolved the fiduciary model of agency for financial services over the past 75 years. It has shouldered the weight of market turmoil, and we continue to refine it as the center of gravity for ethics. The US Department of Labor recently completed a revision of its fiduciary standards that went into effect on 9 June 2017. The US Securities and Exchange Commission (SEC) also made efforts to update its fiduciary interpretation as part of the Dodd-Frank Wall Street Reform and Consumer Protection Act.

The Committee for the Fiduciary Standard mandates that fiduciary agents must demonstrate five key principles:

1. Putting the client’s best interests first.
2. Acting with prudence, that is, with the skill, care, diligence, and good judgment of a professional.
3. Not misleading clients. Provide conspicuous, full, and fair disclosure of all important facts.
4. Avoiding conflicts of interest.
5. Fully disclosing and fairly managing, in the client’s favor, unavoidable conflicts.

As a whole, the fiduciary framework provides a reasonable and measurable set of principles for financial agents to follow.

“Second star to the right, and straight on till morning.”

As with all major inventions that influence the course of civilization, a combination of social, technological, and economic rationales make machine learning and AI irresistible. And the rush for competitive advantage will probably outpace the more careful strides of compliance.

So what can be done?

The financial industry has this rare moment to guide the AI revolution in a safe direction, one that leads to the greatest benefit for all. So how? We should establish an industry-led consortium composed of investment, IT, and machine intelligence professionals to create a practical, evolving and open-sourced framework for machine agency in investment management.

This broad effort would create a central forum for disclosing and mitigating risks to safeguard the health of the financial system, and provide developers of financial machine applications with realizable standards that guard against known and future AI/machine intelligence risks.

AI is here to stay and is developing at an accelerating pace. We now have an opportunity to shape the development of this remarkable technology and usher in a safe era of AI/machine intelligence investments.

If you liked this post, don’t forget to subscribe to the Enterprising Investor.

All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer.

Image credit: ©Getty Images/monsitj