clock menu more-arrow no yes

Introducing Binatix: A Deep Learning Trading Firm That's Already Profitable

The startup's execs say applying the AI technique has given them an edge in investing.

Courtesy: Binatix

While Silicon Valley’s brand names use “deep learning” tools to help computers recognize cat images and play games, one startup has been quietly applying the artificial intelligence technology to more straightforward aims: Making money.

Binatix is effectively a deep learning trading firm, possibly the first to use the state-of-the-art machine learning algorithms to spot patterns that offer an edge in investing. The seven-year-old Palo Alto, Calif., company, which is emerging from stealth mode with the publication of this story, says it’s already nicely profitable.

“It’s been working very well for over three years,” said Itamar Arel, chief technology officer and co-founder. “It’s beyond luck; it’s clear we’ve got an edge.”

The company declined to discuss financials in detail.

Leveraging software, data and quantitative analysis to spot investment opportunities is far from a new phenomenon on Wall Street. And while elegant algorithms have minted millionaires in boom times, they’ve also been implicated in the unraveling of high-flying hedge funds and near-meltdowns of the market.

So at this point it’s hard to say just how novel Binatix’s approach is and whether it will stand the test of more volatile times. But the company’s co-founders believe they’ve made a genuine leap by applying their particular flavor of deep learning to this problem.

What is deep learning? As Re/code explained in an earlier piece:

Deep learning is a form of machine learning in which researchers attempt to train computer algorithms to spot meaningful patterns by showing them lots of data, rather than trying to program in every rule about the world. Taking inspiration from the way neurons work in the human brain, deep learning uses layers of algorithms that successively recognize increasingly complex features — going from, say, edges to circles to an eye in an image.

Notably, these techniques have allowed researchers to train algorithms using unstructured data, where features haven’t been laboriously labeled by human beings ahead of time. It’s not a new concept, but recent refinements have resulted in significant advances over traditional AI approaches.

Unlike most deep-learning approaches to date, Binatix’s software doesn’t just learn from static data points, but incorporates “temporal signals,” essentially how the information continually changes over time, season by season, day by day, minute by minute. That provides them with something closer to a three-dimensional view of financial trends.

Think of it like a still image versus video — you can perhaps detect the general direction of motion in both, but the latter shows the range of motion, speed and other details that suggest what will happen in the frames to come.

“If you have a deep learning architecture that can identify those patterns across different time scales, you can arguably use that information to better forecast what will happen next,” Arel said.

Tech’s biggest players are laying bets on the promise of deep learning across a range of applications.

Google bought DeepMind, Twitter picked up MadBits and Facebook nabbed New York University professor Yann LeCun. Meanwhile, venture capitalists and entrepreneurs alike are rushing to plug money into startups like Nervana Systems and Vicarious (and we hear just about everyone is competing for the rare engineers with real experience in the space).

Arel is a visiting associate professor at Stanford and was previously a computer science professor at the University of Tennessee, where he focused on artificial intelligence. He co-founded the company in 2007 with Chief Executive Nadav Ben-Efraim, who previously worked as a business development and marketing executive at several acquired startups, including Silicon-Spice and Passave.

In 2010, Binatix raised a “small amount” of funding from angel investor Scott Banister, who previously invested in Paypal, Zappos and Uber.

The company initially explored using the deep learning algorithms for speech recognition technology, and found it produced interesting results for most of the problems they tested it on. They realized that if they applied it to stock investing, the company might not have to raise additional capital, Banister said.

It worked.

“The Silicon Valley model is generally you just have to keep raising round after round of dilution on this stuff,” he said. “But we’re basically done. It’s the easiest path from financing to the end of financing that I’ve ever had.”

Technically, Binatix is what’s known as a proprietary trading company. It’s distinct from a hedge fund or venture capital firm mainly in that it invests its own money, rather than funds from limited partners and accredited investors.

As a side business, the company also works with hedge funds, using Binatix technology to develop and implement investment strategies on their behalf. They declined to name any customers at this point.

“We are actually doing pretty well,” Ben-Efraim said. “We are unique in that we’ve created a real business around deep learning technology, beyond being able to recognize cats and dogs.”

This article originally appeared on Recode.net.

Sign up for the newsletter Sign up for The Weeds

Get our essential policy newsletter delivered Fridays.