Watch out, humankind: a brand new super computer from the Massachusetts Institute of Technology has just blown away a good chunk of human competition in a new test.
MIT has been testing a new computer system that is good at recognizing patterns, something humans have traditionally excelled at an artificial intelligence has struggled with. Called the Data Science Machine, it was able to beat a number of human teams in looking through a database of promotional sales dates and weekly profits in a recent test, according to a UPI report.
While computers are far faster than humans at most menial tasks, pattern recognition has been one area where scientists have struggled to figure out when designing new forms of artificial intelligence. While computers can often be very good at spotting patterns, they don’t do a good job of sorting and finding meaning to those patterns. However, MIT researchers think that the Data Science Machine could be the answer.
So they put the super computer against 906 human teams in a recent test. It managed to beat 615 of the teams — not perfect, but far better than previous computing systems.
The task the teams had to take on involved working on an algorithm capable of predicting patterns. The machine was able to do the prediction in no more than 12 hours — and sometimes in just two hours — whereas human teams would typically take months.
Then, the computer looked for similarities in the data by making use of numerical identifiers, and then updating those identifiers as it goes through the data. After that, the process is refined, and the Data Science Machine can start identifying trends.
Despite all the hand-wringing about AI taking over like in futuristic films, this may actually be a very good discovery for mankind. It could alleviate a lot of work, especially jobs that require analyzing data, and therefore free up labor for other uses.
The findings were to be presented at the IEEE International Conference on Data Science and Advanced Analytics.