Deep Learning Playbook for Investment Research
Survive in a highly competitive AI economy.
Much ink has been used on the importance of AI, and more specifically, Deep Learning in the lives of our businesses. That it is the new electricity.
The business of Research is one of the most at-risk to dis-intermediation by Deep Learning (for the purposes of this write-up, I’ll use AI and Deep Learning interchangeably, even though they are not. It’s just that AI is more widely recognized by the populace).
The domain of research is much more than the be-speckled grad student toiling away beside glass beakers in a science lab.
For example, one off the most competitive fields to enter is being an equity research analyst. Equity research analysts on Wall Street can earn as much as $115k+ their first year after college. A sexy job and salary. There is much incentive to becoming a research analyst.
This $115k (going up to the millions, if wall street bonuses are included) is also a cost to the bank. The bank has an incentive to lower costs, say by hiring one less analyst but augmenting the remaining ones with research pre-processed with AI.
In fact, this is what both the equity research and the investment banking departments are focusing on at this moment.
Now, “research’” can be broadly defined as ingesting a lot of data, indexing/arranging the data into a coherent pile, and ultimately triggering some sort of recommendation or conclusion from the research.
This type of research is certainly done by any investment organization (and as we have seen, spending a lot of money on the process). Additionally, within the capital allocation process, the types of data being ingested and researched are varied, as are the industry ‘verticals:’ oil/gas, shipping, retail, tech, you name it.
All along the investment process — from deal origination (‘I’d like to help you raise money, but I need to study your business’) to bankruptcy (unearthing value in bankruptcy document papers — ‘vulture investing’ ) and beyond, data and content are manipulated day in and day out.
All corporations, large and small, do all sorts of research: a Japanese telecom company needing research into corporate venture capital targets in the US, for example. Or a Brazilian beverage company seeking more investment opportunities in Europe.
There’s clearly a need to get ahead of the growth of data sources and the speed at which these data are growing. If the decision is made to get started on AI for research, the next obvious question is, how?
While there has been a flurry of articles from academia about the latest breakthroughs, and some surfacing about other companies beginning to implement their Deep Learning projects, there are very few examples of how to actually get started on a Deep Learning project for research.
There are a few reasons for this.
- The changes are happening so fast. Some might think the correct strategy is to wait until the technology has ‘settled’ before embarking on a Deep Learning project for research. Kevin Benedict, senior analyst at the Center for the Future of Work, writes:
We surveyed 2,000 executives across 18 countries for our Work Ahead report series and they predict AI will be the digital technology having the largest impact on their work by 2020. Even though only 15% of the respondents think artificial intelligence is having a large impact on their business today. Forty-six percent believe AI will be critical to them within the next 40 months — that’s a 207% predicted increase in business impact.
2. Where do I find the AI experts? You might have heard there is a 2 million shortage of qualified computer programmers. There is an even bigger demand for data scientists. We would suspect there is a more acute demand for AI/Deep Learning professionals. Cade Metz of Wired writes:
Microsoft research chief Peter Lee says the cost of acquiring a top AI researcher is comparable to the cost of acquiring an NFL quarterback.
3. Eleven other thoughts. We published eleven other thoughts on why there might be a cognitive gap between the worlds of business and AI, in our piece “Eleven Biases why Experts are Missing the Train on Deep Learning”.
Do you know what? You should get started anyway.
Get started because, like electricity was to businesses in a previous era, AI will change the business landscape for this era.
There are relatively risk-free and non-destructive ways to get started in adding Deep Learning to your research initiatives.
Measure (many times) cut once. The old adage still works in the digital era.
Before you back up the truck at Home Depot and start buying planks of wood, windows and light fixtures, you spend the time and effort to first create your project on paper.
In the building example, you get an architect to design what your building would look like. In the world of AI entry points, this is working to define a Technical Roadmap to building your AI project.
I call it relatively risk-free and non-destructive because no permanent, large-expenses need to be made. No permanent hires, no expensive hardware and software teams. In the era of the Gig Economy, a Technical Roadmap is the perfect entry point to:
a) get you working on the thought processes needed to roll-out a Research Proof-of-Concept
b) make you aware of which of your research workflows can be augmented by AI
c) give you options to pinpoint cost savings afforded by AI-enhanced Research
d) get you to put down your AI books and journals and start working on doing AI
We’ll work on fleshing out your options for getting started on your path to implementing a Deep Learning project, from Roadmap to Proof-of-Concept to Pilot Project and finally Production Environment. All in a relatively risk-free, and non-destructive way.
For more information on Research Machine visit: http://www.intuitionmachine.com/research-machine/
And read our blog: https://medium.com/intuitionmachine
