“Don’t sweat the technique.” — Rakim

Automation of Systems, Augmented Game Play

I wrote this post about eight months ago, but never published it. Today I read @Saeed Amen’s post on “AlphaGo;” DeepMind’s pet project to apply machine learning to a game with a complexity of chess × (1.0 × 10¹⁰⁰), and it reminded me to updated this piece on environmental analysis and automation.

(A snapshot of decision tree processing conducted by complex systems to determine probabilities for in-game success)

For those who haven’t seen it, Timo Boll (big in the ping-pong *ahem* table-tennis world) faced off against KUKA Robotics the other day (now eight months ago) in an effort to exemplify the precision of industrial robotic arms. The video is a little over the top — as are most of their videos — but worth the three minutes of your time.

Thinking more broadly here; there’s a robot for everything these days. One of my personal favorites is a trained neural network which finds the optimal input to beat Super Mario. Now with AlphaGo, we have the same concept, optimizers for games.

(In the game (link below) Mario finds the optimal route for completing levels thanks to a neural networks)

Strangely, most of these optimizers die. Or at least, they remain in-game optimizers and go no further than the game themselves. But where else can neural networks be applied?

Different industries have different rates of technology turnover, usually associated with the industry talent turnover: the lower, the lower. But for some industries, the mandate IS technology innovation, changing the speed at which new forms of processing are implemented and distributed across divisions and teams.

Financial services seems split between which one of these roads it wants to take. Be a utility service, with low turnover and little technological advantage, or be constantly re-inventing itself and assure technology innovations are top of mind.

For us, we prefer the latter, as that’s how we come into the picture for the companies exploring emerging technologies. Without an industry mandate to increase the number of factors they apply to whatever models are being built, there’s no market for the stuff we’re building. If however there are forward-looking companies that understand why they might need access to more than just price and fundamental data, then we are indeed in the first inning of a MUCH larger game to come.

“We’ll hire you to automate yourself out of a job!”

If we continue at the current rate of active search for technological edge we’ve begun to see in the market today; in the near future, traders — as classically imagined — will really become systems engineers. With the holy grail being fully autonomous systems, trained by neural networks. I am an advisor to two companies who will hopefully make this easier for every type of investor or trader in the future.

With hardware, software, and data as building blocks, developing systems that can beat other systems is the new game, and beating other systems in financial services, means more than just sweet YouTube videos.