Investing in AI: Balancing Diversification with Specialization
Written by Humza Saleem, Strategist and “Humza” at Hypergiant
Warren Buffett famously stated, “Diversification is protection against ignorance. It makes little sense if you know what you’re doing.” In his view, investing in a few industries well-known to the investor is more lucrative than spreading one’s resources across multiple, novel industries. And while his take is contrarian (but sensical) and runs counter to traditional portfolio theories (most of which suggest that investments should be diversified across many sectors and/or asset classes), forecasting success with infant technologies in a new arena is nearly impossible.
Effective investing, particularly at the venture capital (VC) level, requires a delicate balance of risk-tolerance and risk-aversion — an equilibrium of specialized, domain knowledge with traditional portfolio diversification. Such a hybrid investment strategy is intended for a very particular type of animal and, in this case, it’s Artificial Intelligence (AI).
AI is a highly dynamic industry that fosters a confluence of buzzwords and startups establishing businesses just for the sake of doing so. And yet, it also provides clearly defined customer value. It requires closely monitoring merger and acquisition (M&A) activity and critically identifying the business problem at hand, while simultaneously heralding an appetite for taking risks in a world full of imperfect information.
Why Should You Care?
In the last decade alone, there has been a 14x increase in the number of active AI startups, a 6x increase in funding to AI startups by venture capitalists, and a 4.5x increase in the share of jobs requiring AI skills. What started off as basic, “rules-based” application design has turned into an open field of business opportunity that includes everything from predicting medical diagnoses to forecasting the movement of financial market indices. And, thanks to the overwhelming research and acquisition interest by major tech brands, it’s a technology you likely won’t stop hearing about anytime soon — and investors are keen on quickly getting a piece of the pie.
For context, in 2017 alone, VCs around the world collectively poured more than $11 billion into AI and machine learning startups, contributing heavily to the huge increase in global VC financing since the beginning of the decade. There were nearly 367 equity deals, more than 1000 unique investors, and 11 unicorns (denoted by their $1 billion and above valuations) last year alone.
Catalyst(s) of Opportunity
Before exploring how AI impacts industry, it’s helpful to take a step back and understand how AI has changed over the past few decades.
First, we’re collecting a ton of data (albeit in a disparate, fragmented fashion) at a much faster rate than ever before. Research suggests that by the year 2020, about 1.7 megabytes of new information will be created every second for every human being on the planet. This collection of data about people as users, employees, and customers creates building blocks ripe for artificially intelligent solutions.
For most organizations, however, this accelerated advancement is a pain point. Harvard Business Review reports that the biggest obstacle to leveraging data as a competitive advantage isn’t skill base or technology — it’s plain old access to relevant data. And, although companies are collecting more data than ever before, most don’t how to use it or what to use it for.
Second, as the operational costs of computing power and storage continue to drop, the capacity of modern-day technical processes steadily grows. According to AI Impacts, computing power available per dollar has increased evenly by a factor of ten roughly every four years in the last quarter of a century, a phenomenon related to Moore’s Law. This suggests that, over time, we might witness an inverse relationship between the computing costs and computing power of AI.
Last, but most important, we’re seeing tremendous improvements in deep-learning methods. These networks not only pose durability and scalable value, allowing for richer exploration and correlation as data grows, but they also aid us in identifying the biggest opportunities for AI solutions. Thankfully, this has kickstarted a community-wide conversation that prompts a better understanding of AI as a value-enabler, while also giving entrepreneurs and technologists the tools and frameworks they need to build profitable businesses around a variety of complex tasks (both at consumer and enterprise levels).
Fusing Opportunity with Industry
The aforementioned developments ask that we move past buzzwords and think critically about industries ripe for disruption. Market analyses must be conducted to uncover where dollars are being allocated and which use cases early-stage companies, digital transformation executives, and venture capitalists are focusing on. Here, we select a few.
The swath of cyberattacks and data breaches in 2017 resulted in an increased interest in combative, AI-based solutions. From the Equifax breach that compromised nearly half the United States to global ransom campaigns that cost organizations millions of dollars, cyberattacks highlight the alarming vulnerability of our personal data. With nearly $1 billion invested in this space in 2017 alone, the opportunity for entrepreneurs to sweep in and save the day (while making some money) is evident.
Take CrowdStrike, for example, one of the few unicorns in the niche field of “Endpoint Detection and Response” (EDR) — an emerging category of tools and solutions for detecting, investigating, and mitigating suspicious activities on hosts and endpoints. The company is considered to be a leader in the threat-intelligence space, has amassed nearly $281 million in Series D funding, and (by some estimates) is valued at $1 billion. Or consider Boston-based Cybereason, another EDR platform close to reaching unicorn status. They’ve raised nearly $200 million in Series D funding from the likes of SoftBank Group and are valued at over $900 million.
AI will redesign healthcare completely — and for the better. From helping medical professionals architect treatment plans to assisting them with repetitive, manual tasks, AI lets medical professionals concentrate on the “human” aspects of their jobs. And, despite barriers to entry and a regulatory environment, healthcare startups raised nearly $750 million in funding last year. They tackled problems across imaging and diagnostics, virtual assistants, remote monitoring, in-hospital care, and more, propelling a record rate of new companies every quarter.
Flatiron Health, which uses human-assisted machine learning to mine health data for cancer patients, is a healthcare unicorn. After raising $175 million in Series C from the likes of First Round Capital, Google Ventures, and SV Angel, the company reached its billion-dollar valuation in late 2016 and, in early February, got acquired by Roche for $1.6 billion. This is an example of how a small company can solve a big problem using technology, all while having a clear exit strategy.
In a market dominated by the likes of Amazon, Google, and Apple, startups can leverage their speed and agility to tackle enterprise-level problems. A concrete example of this is automating routine tasks and processes. The ongoing debate as to whether AI and task automation create jobs or destroy jobs shows both camps missing the bigger picture: automating routine tasks will help workers get better at their jobs, learn skills that add value to themselves and their companies, and (ultimately) benefit their employers’ bottom-lines.
In wealth management, for example, companies like Wealthfront, Betterment, Fount, AIM, and Sentient have created robo-advising capabilities, collectively raising nearly $1.32 billion in funding since 2012. This enables optimal customer experience as the human advisors behind the scenes tackle more personal, sophisticated tasks with their clients.
Human resources (HR) is another area with an abundance of data and need for AI algorithms. Google recently released a beta version of its machine-learning-based Cloud Job Discovery platform. It uses neural networks and natural language processing to map job descriptions to candidates with the right skill sets — something traditionally done by recruiters.
AI also has enormous potential to improve efficiency in legal environments by simplifying the research- and time-intensive aspects of law. ROSS Intelligence, a venture-backed startup built on IBM Watson, has launched a platform that allows lawyers, paralegals, or anyone else researching legal cases to ask questions in natural language and receive answers that are based on prior proceedings.
What Does This All Mean?
Startups rise and fall. For every billion dollar company, there are nine that failed. As Jake Flomenberg, partner at Accel Partners, puts it, “Most AI startups are not going to fail on the basis of their actual AI, but fail because they fail to identify a problem that needs solving.” As people that live and work within the realm of technology, we shouldn’t build platforms or businesses simply for the sake of doing so. Instead, we should consider the fundamentals of taking a “problem-first, technology-second” approach to business creation.
The same mindset should be taken by investors, too, when considering an emerging technology. But this alone doesn’t guarantee success — imperfect information, outliers, and exceptions will always exist. So while we agree with Buffett’s suggestion to specialize around a few sectors, we also know that risk remains. Which is why creating an investment portfolio that’s both diversified and specialized seems like the appropriate way to go. Because AI is here. What remain to be determined are the problems worthy of the technology, the innovators building the right solutions, and the capitalists making it happen.