X
Innovation

How to augment our intelligence as algorithms take over the world

Small, but very powerful algorithms are ingesting data and making decisions at a rate that is far faster than any human brain can comprehend. How can we keep up with the machines?
Written by Mari Silbey, Contributor

Physicist, mathematician and Chief Technology Officer of Quid Sean Gourley will paint a picture of the not-too-distant future in his talk this weekend at a TEDx event in Mountain View, California. That picture includes a whole new ecology of beings popping into existence, fighting for survival and supremacy, and finally exhausting themselves under the inevitable force and weight of evolution. The beings that Gourley will talk about, however, are not living, breathing entities. They’re algorithms.

I caught up with Sean Gourley this week to hear his thoughts on the ecology of algorithms, predator/prey relationships, and what we as mere mortals can do in the face of computers that move at speeds far faster than our brains can perceive. The conclusions he presents are both terrifying and endlessly fascinating.

First, despite all the talk of Big Data, Gourley sees a more important trend in the rise of small, fairly dumb algorithms that happen to be very powerful. These are programs that follow information, determine patterns, and use trigger points to set new sequences of activity in motion. Right now, algorithms already control the majority of equity trades in the US financial market. Humans loose these algorithms into the wild, but they move at such a rate that we can’t follow them or even regulate these programs before they’ve moved on to the next level of market analysis, and the next batch of trading decisions.

Ecologists are now tracking financial algorithms and finding that their behavior mimics that of predator and prey animal species. One program may try to hide a large transaction, for example, by creating swarms of smaller transactions meant to fool a competing program. Sometimes the trick works, and sometimes the competing program detects the subterfuge and counterstrikes.

Beyond financial markets, the same trend is starting to take hold in the advertising world, with algorithms determining when and where advertisers should bid on available inventory. There are also retailers who use computer programs to set prices on Amazon, and a wide variety of recommendation engines relying on algorithms to decide what information should be presented to us for review.

So, should we be worried? Are the machines taking over? Is a Skynet-like authority inevitable? Gourley says there will likely be a backlash as more people realize the control we’re rapidly relinquishing to computer programs. But he also points out that a lot of things these algorithms do, aren't things we want or have the resources to do ourselves. Take GPS software. Increasingly GPS software analyzes not only maps, but also construction work, traffic congestion and more to determine the fastest route to any destination. Perhaps we could scan the same data sources to map out our own path, but it would take us far longer, and probably wouldn't be more effective.

Gourley suggests that the best answer to understanding and regulating complex data systems is to start augmenting our own intelligence with computing tools that make these systems human-readable. In other words, we need to create layers of abstraction so that we can fathom the activity taking place, even if it’s happening too fast for our human brains. Gourley’s company Quid designs some of these tools, creating multi-dimensional models to help people understand vast quantities of ever-changing data. Humans, he says, can manage three-dimensional models, handle the addition of a fourth dimension – time – and even comprehend a pseudo fifth dimension through size depictions of changing data conditions. (There may be a viable fifth-and-a-half dimension too in the use of granular color changes to represent different types of activity.) Once we hit six-dimensional models, however, Gourley says humans will struggle. Our sweet spot appears to be right below that level.

With sophisticated modeling tools – even up to only five and a half dimensions – Gourley believes humans can stay in control of machines. Or, at least we can point our algorithms in the right direction. Just because the world is growing more complex, he says, doesn’t mean we can’t still make decisions. We just might need a little extra help.

This post was originally published on Smartplanet.com

Editorial standards