Data Science Machine — MIT Creates Computer That Outperforms Human Intuition

Artificial intelligence.

The quest to build an artificial being that can do, act, and think exactly like a human being has long been a pursuit of mankind. The ability to play God, the ability to reproduce ourselves in an artificial sense is a notion that has long smoldered in the subconscious of the human species. Long before the conception of the modern computer, artists, writers, and filmmakers have toyed with the idea of an artificial man. From Mary Shelley’s roots in Frankenstein, to Fritz Lang’s depiction of an artificial human in Metropolis, all the way up to the most popular movies of our own generation with examples like C-3PO in Star Wars and Ultron in The Avengers. In the real world, Siri and Ask Google can sometimes make us feel like we have an artificial human in our hands. There’s no doubt about it: mankind is obsessed with artificial intelligence.

Ask Google

However, though we’ve been able to produce robots and machines that can talk and act like humans, though we’ve built machines that can interact with humans to a large and convincing degree, we’ve never broken the barrier of artificial intelligence. We have never, up to this point, created a being that can consciously think and reason on its own without the aid of humans — but we’re certainly getting closer.

Researchers at the Massachusetts Institute of Technology have come up with something called a Data Science Machine that they say can outperform human intuition. In the past, computers have been able to outperform humans in many tasks, including information retrieval, processing, and even data sorting. However, what programmers of computers have never been able to replicate in the past is a human being’s ability to recognize not only patterns, but to infer intuitively as to what the pattern indicates. Human beings alone are able to take in information and think “outside the box,” if you will, as to what a certain set of information can imply outside of the parameters of the given information.

The researchers at M.I.T. now say that they’ve changed all that.

The test that the scientists at M.I.T. is a little complicated. Basically, what they did was test the Data Science Machine against 906 human teams in a pattern implication competition. Both the Data Science Machine and the humans worked on an algorithm for months that would predict complicated patterns. When the results of the tests were compiled, the Data Science Machine came out on top, beating the humans in 615 instances out of the total 906. During the tests, both the humans and the Data Science Machine had to discover correlations in a large data set using numerical identifiers. The identifiers had to be continuously updated as the data set was compiled. Once the data was successfully compiled, the humans and the Data Science Machine had to figure out patterns contained within the data, and as a result they had to predict future patterns and trends.

So what does all this mean? What could the Data Science Machine do in a practical sense? One of the examples the team at M.I.T. gave was that the machine could examine a particular online class and predict the percentages of students that would drop out. That seems pretty anomalous and a bit vague. Perhaps in a larger sense, the Data Science Machine could also accurately predict trends in the stock market, predict the likely behaviors of foreign governments, or correctly guess the outcome of the human experience — warning us of our impending doom or success.

Then again, it could go AWOL, decide that humans are inferior, replicate itself, take over the internet, enslave mankind, and become its own God.

"Sir? We have a problem... the robots just shut down the internet."

The M.I.T. researchers will present their findings at the IEEE International Conference on Data Science and Advanced Analytics.

What do you think about artificial intelligence? Should humans be messing around with things like the Data Science Machine, or has the pursuit of A.I. gone too far?

[Photos by Oli Skarff, Jeff J. Mitchell and Carl Court / Getty Images]