October 2, 2016

Collected Intersections: AI and Money (It’s just a number, after all)

These days, I’ve been asked a handful of times to chat about artificial intelligence and my perspective on how it intersects with some problem or field. In this series of Collected Intersections, I’m highlighting a handful of links I’ve come across that tie the current topic, AI, with money.

First, a few brief words on how they tie together.

Algorithms – and ‘artificial intelligence’ is a catch-all to describe particularly clever algorithms – exist to pull meaning from noise, and to do so over and over again. The ‘state of the art’ of these algorithms, and the agility of the machines that can run them, have advanced to a white hot degree over the past few years.

Money needs no introduction, but a few notes might help frame this Intersection: modern money exists (mostly) in computer-land, as sets of numbers in a bunch of databases. Money ‘changes hands’ less often than it moves back and forth across rows and columns.

In theory, all of the companies that have a hand in moving money around (or anybody else observing the movements) could take snapshots of how money moves, curate those shots, and use those pictures to understand some deeper truth. With deeper truth comes asymmetric knowledge, and with that knowledge comes the opportunity to profit.

That’s the model of every company that collects data as a business goal: the more you know, the more you can predict, and the more dollars you can catch with your wallet before everyone else.

So… What’s the problem? Why is it hard?
First, you need to 1) figure out what you want to predict, and why, and how it will help you make money.
Then, 2) build, buy, or borrow a way to collect relevant data.
Then, 3) collect the data.
Then, 4) curate it.
Then, 5) mechanically sift insights out of the data and understand it.

By this point, you might have asymmetric knowledge, if you’re the only one collecting at step 3, or the best curator or sifter at steps 4 and 5. And, if your step 1 works out, then, you can finally begin to leverage that asymmetric knowledge for profit.

So, if you have a good, novel business idea (and good luck), the hinges are really at collecting data and sorting data. This applies to stock markets, car-sharing services, Amazon-type marketplaces, and any other instance where markets have fluctuating prices.

Rogue Machine Intelligence and A New Kind of Hedge Fund
June 21, 2016
Hedge fund decentralizes the sources of bets (anonymous machine learning agents) creates ensemble learner. Twist: pays anonymous contributors in bitcoins, also allows contributors to donate to an AI safety net organization.
AI in Digital Wealth mgt: Algorithms
March 23, 2016
An overview of the startup scene in financial industries, and how they’re using algorithms and data to make money.
A Survey of Deep Learning Techniques Applied to Trading
May 30, 2016
An overview of the academic world: algorithmic techniques used to predict market movements in very specific cases.