Financial singularity and the death of capital markets
Financial singularity is considered as far less dangerous than technological singularity because artificial intelligence, when applied to finance, doesn’t directly threat our physical integrity.
Harmless.
Is it ?
When the brightest minds in artificial intelligence and robotics warn us about the threat offensive autonomous weapons represent to humanity, we realize that technological singularity could be imminent.
High-profile influencers and researchers with the names of Stephen Hawking, Steve Wozniak and Elon Musk tell us that the global AI (artificial intelligence) arms race may start soon, or, worse, that it has already started. They tell us that we, humans and the people who lead us, have the responsibility to stop this before it is too late.
Critics of technological singularity argue however that such warning signs are unnecessary because we are at the embryonic stage of military artificial intelligence. The basis for their argumentation is that similar warnings were issued 20 years ago over cloning, perceived as a major threat to the sustainability and the evolution of our human race. Unnecessary warnings because no real threat has arisen from cloning experiments. Out of context, this may be true. But we are in 2015 and technological breakthroughs supported by shorter discovery and development cycles allow for technologies to progress much faster than 20 years ago. And to have deeper impacts on our lives. Unarguably James Cameron’s 1984 Terminator was science fiction at the time it opened in theaters and remained science fiction for 30 years. Now to become realistic and possible. AI-powered human-like robots could see life soon and we can express doubts about the willingness and ability of our leaders to stop this. Musk, Hawking, Wozniak. Against Moore’s law and corollaries, who stands a chance ? Things are moving really fast. People leading change and disrupting industries today are not the researchers experimenting cloning on sheep in the 90s. They are entrepreneurs with unprecedented financial reach. Rogue nations may not lead the AI military race or even trigger the need for this. Billionaires running the next unicorns or decacorns may do.
Whether mankind will stand against offensive autonomous weapons before technological singularity and will stop further research in this space or not will certainly be at the center or many debates. And we are just witnesses.
In the world of finance, the topic is overlooked and the debate quasi-inexistent. Financial singularity is considered as far less dangerous than technological singularity because economic weapons, if they exist, don’t directly threat our physical integrity. The automation of banking processes, robo investing software, AI-based black box hedge funds, crypto-currencies and digital banks are considered harmless. That we think they are. But financial singularity may do more harm to our societies than offensive autonomous weapons, only because the latter would be intelligent enough to know when peace is the best solution for all. To leave a role for humans. Financial singularity, on the contrary, may simply lead to removing the element that makes financial markets inefficient : men.
The perceived mirage of financial singularity
Several economists including Robert Shiller talk about the mirage of financial singularity; the impossibility for it to happen because of the major impact human behavior has on the financial markets. But Shiller only touches the surface of what total disruption of our financial markets could be. And to some extent this is an insult to the potential of AI. Full power in AI does not lie in analyzing core economic and market data, it does not lie in unbiased investing or in the speed at which algorithms can gather and process information. It does not even lie in learning how men trade and how participants behave, or in accounting for irrational patterns. By learning, understanding, anticipating and eventually mastering concepts such as greed & fear, panic selling, moral hazard and other behaviors, algorithm-based advisors will quickly be able to neglect human intervention. What we, investors, do today to the markets will just be noise and AI systems will take over.
The Chinese stock market is under pressure. Some say it is a healthy correction. Economists, financial analysts, fund managers have warned us about the heat in the market and the bubble, for months or years. Would AI investment vehicles have allowed to cold down the market before the recent sell-off ? Yes, but only if AI-induced flows had surpassed traditional flows. Only if Chinese and overseas investors had agreed to or wanted to consider what AI technology was telling them. So the only answer is no. Not in 2015.
In the airplane, the pilot is a problem. And the right the pilot has to disconnect the autopilot is unfair. Why can’t the autopilot have the right to override a pilot’s action. Germanwings. The computer knows. Humans don’t and influence systems the wrong way. We don’t want to allow robos to drive the markets, even if they are more able to do this than us.
So Shiller is right. Financial singularity is a myth because men influence the market. Because markets are not efficient and news, flows and behaviors have a deeper influence on stock prices than rational factors could ever have. Missing or exceeding analyst expectations drive prices more than the reported financial data themselves.
But financial singularity goes well beyond the idea that AI could meet human behaviors. It goes well beyond automation. Most of the processes in retail banking can be automated, since humans are only adding risks for errors instead of value. No AI is even needed to create retail digital banks. Basic scripts can mimic most, if not all, operations. And do better.
In the investment world, robo advisors are gaining in popularity because they demonstrate that the value added by most investment managers to their clients is quasi-inexistent. Even of greater importance, they offer solutions at a fraction of the cost. Low cost beta investing and simplified asset allocation for the masses are attracting large family offices, UNHWI and institutional investors. But in this field, again, we are more talking about automation, fast processing of data and rational investing rather than AI-investing. Translating one investor profile into an asset allocation map and investable products, mainly ETFs, is no rocket science. Portfolio optimization not even more. And robo advisors don’t even come close to override investor behavior flaws.
While this is the situation today with Shiller’s case being valid, I argue that tomorrow’s situation may significantly differ. That financial singularity is not what some anticipate it to be. If we agree that progresses in AI R&D will continue to accelerate, and accelerate even exponentially, both in development cycles and capabilities, then we should revise our very own understanding of financial singularity. Way beyond robos understanding participants’ investment and trading behaviors.
Algorithms can easily identify behavioral patterns when they try to anticipate market movements driven by activities built on irrational grounds. We now have access to AI systems that can trade using fundamental analysis, that can include and analyze flows, analyst opinions, irrational behaviors and all sorts of patterns. Some systems can and will self-improve to reach a first point of non-return, that shall triggers many alarms, increase market risks and volatility. By learning how markets move and how irrational behaviors trigger flows greater than the fundamental ones, machine investing will align with human investors. To even precede and trigger worse behaviors. Schiller’s point is that in case of a massive sell-off, irrationally-triggered flows will dominate, increasing the downside even if machine advisors would recommend buying. Automated systems will be able to anticipate such market movements, knowing ahead of time that human will sell. Machines will trigger the correction. Anticipation and prediction will be based on the analysis of unstructured data coming from all sources, building predictive patterns indicating that market participants are on the verge of selling. The most sophisticated systems will analyze market data, trading flows, research, internet searches, social media, specialized sites, but also combine the analysis with financial institutions’ own databases and systems. Adding an extra layer of information, such as geolocalisation, ATM activities, CCTV, satellite images, machines will know everything before market participants act. Human will simply do what they intended to do, but not just by following a trend, but by following AI trading activities.
This first phase will allow an alignment and an overlap of AI-induced flows and human ones. Algorithms will rationally include human’s irrational behaviors to improve their predictive power.
Point of no return
Because then, trading flows generated by AI systems will be big enough to influence the markets and surpass human activity, not just influencing it. Schiller’s case is losing grip. And financial singularity is still nowhere near. But the steps towards it are set.
When AI flows coincide with human trades, AI-based systems will have achieved two things: first, they will be able to set a fair price on financial instruments and assets, using multiple factors and degrees of optimization; second, they will have learnt the keys to the Keynesian beauty contest, mimicking human behaviors and anticipating irrational flows that can and do move markets.
After this phase of high volatility, when accentuated and leveraged irrational behaviors send markets to Heaven and Hell, will come a phase of managed markets. One could eventually call the market genuinely efficient. Robo-advisors, AI-managed funds, rebalancing tools, portfolio management systems will all apply these unprecedented capabilities to not only predict markets using deep understanding of human behavior, but will start directing the market to fairer valuations. The golden age of efficient markets. A haven for fundamentalists and adepts of fair pricing. No financial or asset bubbles. Low volatility. No market manipulation. No surprise.
Boring. Boring, because investors can benefit from inefficiencies, some asset managers (in fact, very few in a consistent manner) are even generating excess returns, alpha over market performance, when capturing opportunities created by market moves, price adjustments and mispricing (read Sean Park’s “(In)efficient markets” on this topic).
AI-driven efficient markets will assure fair pricing of financial assets. Equities, bonds, currencies. But also real assets, private equity holdings or any security that can now be marketed-to-market or evaluated and priced using advanced algorithms that include all rational and irrational models. When humans rely of DCF, multiples of EBITDA, comparable companies or very subjective models to value private holdings during funding rounds, AI will include all of them as well as new models, not yet available or used by us in 2015. Models we may not even understand or be able to approach. With efficient markets will also come a full set of predictive money management solutions. If rationally-priced assets is now the norm and humans can’t influence the markets, portfolio management will become obsolete. Asset allocation will be an automated process. With AI-managed markets will also come the end of financial professionals. Fully automated processes will replace traders and brokers. Fund managers, with the impossibility of generating alpha, will disappear, replaced by ETFs at first, then by new forms of pooled investment vehicles. Ironically, the whole concept of managing money thru investments will be brought down to a very simplified process, but a process managed by over-fitted black boxes. Up to a point that it will become a self-infused and always improving process, non-understandable by humans. 5% per annum ? Sure. 10% ? Sure. Independently of risk profiles. Risk will become totally manageable. Irrelevant concept.
AI algorithms will be able to automatically match needs with a virtual demand and with capital. Algorithm will automatically implement the right capital structure for a company. The right price for new stocks, the right level of debt, the best yield. A new issue will automatically be allocated to the investors who should buy it. No more road shows or pitchbooks. No more research and analysis because humans won’t have time to publish or read them. Stock analysts will have a second life as historians of the markets.
Asset allocation will then disappear, with AI merging asset classes into new concepts. Capital markets will disappear. Bankers will be gone.
The very concept of stock exchange (or any exchange) would become irrelevant, with AI allocating resources to needs directly.
Pension funds will be gone. From a salary, money (assuming this concept persists) will automatically be allocated to pooled solutions to manage future needs and generate returns. Even crypto-currencies will be short-lived. New solutions will emerge allowing humans to get access to goods, according to a conceptualized financial health. Get the vehicle you need according to what you can afford and get it financed the best way.
Capitalism 3.0 with no capital
Financial singularity is about the dematerialization of our banking system and our capital markets. It is about AI redesigning existing processes, developing and implementing new ways of building and managing wealth, in a less Keynesian way. Capitalism 3.0 with no capital.
Financial singularity can happen, simply because at some point AI will surpass the first barrier, identified by Schiller as a hard obstacle to allow this to happen.
We, human, can’t and won’t stop this.
What we need to do it to develop an approach towards this. To already design master frameworks for the banks and the markets of the future, like for directing ivy wines on a wall, we should direct AI in the development of future technological solutions. Think like AI systems. Become AI.
Crypto-currencies, blockchains, robo-advisors, they are just small bricks in a great wall that is being built. Banks and market participants must understand this and have to invent the models of tomorrow or at least anticipate them.
Learn and master them.
They have to become proactive and far less reactive, and focus on the bigger picture, 2025+, not on the themes of yesterday or today.
Quite a challenge.
\\ — — August 2015 — -\\
Richard Joye is founding partner at KCHK, a financial advisory boutique headquartered in Hong Kong with global network. Richard and team focus on disruptive technologies.