Let the Machines Do the Learning

As we celebrate Labor Day this month in the US, it is worth reflecting on the past, present and future of the labor force and its role in shaping modern society. The labor movement really took off in the 19th century along with the industrial revolution, where one begat the other. Today, we are clearly in the midst of another revolution beset by advances in intelligent and affective computing.

How can a post-labor society create a sense of identity when the connection between labor and our identity is so strong? Fred Wilson sees this as a challenge and an opportunity, pointing that one of the answers is to offload tasks to automation and machine learning so that humans can address other challenges that require critical thinking abilities.

Companies, from Spotify to Tinder are using affective algorithms to learn user behavior and present them with individualized, contextualized experiences. Over time, these models become smarter as they identify and improve (not unlike a simple human) their capabilities. As they learn to overcome bottlenecks, human input becomes less and less necessary.

In the recruiting industry, the process of hiring the right employee is fraught with inefficiencies that result directly from the limited cognitive capacity that recruiters have. The three elements they are given--the CV, cover letter and job application-- offer an incomplete profile of a candidate and almost no insight as to whether they will meet the current. If the hiring company is in a fast-paced environment- say, the high tech world- anticipating future needs is an even bigger challenge.

Given the grave consequences of hiring the wrong person (and the potentially unlimited upside of hiring the right one), the hiring process cannot be left to humans alone. Relink saw this issue and put together an algorithm that combs through massive data sets of applicants and uses preset criteria to match them with certain recruiters and companies- similar to what Tinder does with matches. A smart algorithm has an unlimited capacity for identifying trends and providing accurate job recommendations and while it is contingent on candidates accurately self-reporting, early tests indicates that it is more accurate and efficient than humans at matchmaking recruits.

More than a decade ago, similar uses of machine learning were being applied to the world of finance. Jaffray Woodriff, founder of QIM, manages an asset base of $2.5 billion through a data science approach. The fund uses predictive modeling techniques to systematically make investment decisions. Their returns speak for themselves; QIM has achieved a net annualized return of 21% over the past 8 years, making the fund a leader in the CTA fleet.

Human behavior is at the crux of the algorithms that QIM and others use; their models are able to pick up on these behavioral patterns and seek to predict what's going to happen in the next 24 hours. While factors affecting prices change around the clock, human reactions to them change in certain subtle ways. The 10,000 mathematical formulas that make up their algorithm, use machine learning each with "flavors of favorite quant inputs, orthogonal to data points like volatility and price momentum" to spot and anticipate patterns in various commodity prices.

The true beauty of technology is that it truly allow us to do more with less; rewriting the plan of the world and changing long-held paradigms. Machines are not going to take over the world- they will just shake up every conceivable aspect of it. While for the foreseeable future, deep-learning machines will remain pattern-recognition engines, the potential for them to learn and improve upon themselves is endless.

With that, us wonderful humans will have much more time to dedicate to important activities, be they high-level thinking or spending time with loved ones. With the advent of such forces, it will truly be possible to ruthlessly guard out time; a finite resource that money cannot buy more of. Thankfully, it can buy servers and machines.