“Alea iacta est” (“The die is cast”)–Suetonius
When Larry Fink, CEO of BlackRock told the New York Times that his firm had to “change the ecosystem — that means relying more on big data, artificial intelligence, factors and models within quant and traditional investment strategies,”
I was pleasantly surprised. When his colleague, Mark Wiseman, explained that the reason for this change in strategy was that “the old way of people sitting in a room picking stocks, thinking they are smarter than the next guy — that does not work anymore,” I was giddy.
Several hedge fund managers have previously indicated in their usual vague ways that they were going to incorporate artificial intelligence (AI) in their investment processes. But this was BlackRock, the largest asset manager in the world, explicitly admitting that our long-held way of generating alpha is failing us, and AI holds the promise of a better future. Most importantly, BlackRock made clear it is not simply going to use AI to improve a few existing strategies. Instead, it plans to transform its entire business to support a pivot to AI.
This was the reason for my giddiness: from this point forward, every other asset manager must either defend their current approach to active management or follow BlackRock’s lead.
A defense of the non-AI status quo requires a torturous argument tainted with self-interest, excuses, hubris, and blind hope. Mr. Wiseman seemed to anticipate this maneuver when he persuasively told the Financial Times, “We don’t believe that hope is a viable strategy.”
Now, when asset owners and consultants ask active managers if they include AI in their investment processes, “no” is not an acceptable answer. Emulation has become the only defensible position.
Emulation in the asset management industry usually reduces to one output: throw money at the problem. For once, this is not an entirely inappropriate response.
But asset managers – especially traditional multi-asset shops with downtrodden active management groups – might struggle with where to throw their money in what Bloomberg has termed “an escalating technological arms race.” Being in a particularly magnanimous mood, I thought I’d offer a blueprint.
To begin, asset managers need to acknowledge that they are not trying to solve a product development problem. Instead, they’re trying to solve a business problem: how could AI help create a more sustainable, competitive business?
So they need to think holistically: Forget about creating a new AI-enhanced investment strategy and instead focus on creating a platform – specifically, a state-of-the-art AI platform that is agnostic as to the investment problem to be solved and that would support advanced machine learning techniques and naturally grow to absorb different traditional and non-traditional data sources as inputs.
Creating such a platform requires skills asset managers currently lack. Thus, be prepared to throw money – a lot of money – at a lot of people: a chief science officer, machine learning experts, data scientists fluent in finding, vetting, and wrangling large amount of real-time, unstructured, noisy data; and engineers to set up an efficient computing environment. Scientists can be a prickly lot with almost tribal affiliations, but under the right circumstances – think Manhattan Project – they can coalesce into a high-functioning team.
Let’s assume that, at least initially, AI will supplement but not replace human investment professionals. My experience is that scientists and traditional investment professionals speak different dialects of the same language. To maximize the contribution of both sides of the house, the firm should hire cross-functional professionals fluent in both AI and investing to serve the necessary translators between these two groups. More broadly, these professionals will serve as AI advocates internally and external evangelists with clients, prospects, and consultants.
Will this blueprint work? Along with my own AI investing experience, I can point to Millennium Management’s incubation of WorldQuant (c. 2007) as proof-of-concept.
While not an exact analog, its orientation aligns with my blueprint: With Millennium’s support, WorldQuant assembled a senior multi-disciplinary team of AI scientists that built an agnostic AI platform, with the Willy Wonka-like name of the “Alpha Factory.” Other team members were hired to procure new, orthogonal data sets to serve as inputs for the Alpha Factory. The results appear robust: According to the Wall Street Journal, the Alpha Factory has churned out four million “alphas” to date and is aiming for 100 million, which Millennium has used to enhance its own human investment decisions (although a new arrangement will allow WorldQuant to run non-Millennium money).
Certainly, there are substantial barriers to implementing this plan or any other like it. But action must be taken. (Let me warn asset managers against throwing money at an outsourced approach: hiring a general purpose machine learning company—a “machine learning for everyone” platform – is not an alternative.) With BlackRock’s pivot, we have crossed the Rubicon. There is no turning back.
Angelo Calvello, PhD, is the founder of Impact Investment Partners and co-founder of Rosetta Analytics, Inc. This essay originally appeared in the May issue of Institutional Investor.