Wednesday, March 30, 2016

How Machine Learning is Taking Over

While speaking with the Times, Urs Hölzle, an executive from Google cloud, stated that he believes that the business of renting out machine learning capabilities will take over Google’s advertising revenue one day. Meaning, that he is predicting that Google Robots will be bigger than Google search, which brought in $16.4 billion in profits last year. 

Machine learning is a new branch of computer science where systems have the ability to train themselves in pattern recognition, or learn things for themselves. Because they can learn on their own, they do not require a programmer to teach them what works and what does not work.

In October 2015, Google created AlphaGo, a robot that mastered the game of Go, a game created in China 2,500 years ago. The game begins with an empty board, two players using different colored stones, which alternate placing stones in squares while trying to grab territory while avoiding getting their pieces taken. AlphaGo learned the game of Go itself, and just beat the human champion in the seemingly impossible to master game. Techinsider states that in the game of Go, “there are more possible game states than atoms in the universe.”

Many companies are on a quest for machine learning products—which could take over many jobs that are currently completed by humans. Industries as diverse as advertising, energy, cars, medicine, and finance are all being affected by machine learning. Already, at Goldman Sachs, the responsibilities of $350,000 a year jobs are being automated.

Although the Techinsider article did a great job of explaining what machine learning is and the amazing technological advances that have surfaced because of it, it failed to mention any cons to machine learning. This technological advance could be beneficial in some aspects, it would exponentially increase the unemployment rate, leaving more and more people without jobs every year. This advance could be detrimental to the economy, as the machine learning devices are expensive to produce, and so many people will be out of jobs. The article also did not state how much it costs or how long it takes to create products like the AlphaGo or the driverless car. Do the machines learn things quickly, or did it take years for the AlphaGo to master the game of Go?

1 comment:

  1. Although technology advances can be really useful and innovative, I think that trusting machine learning products with $350,000 a year jobs seems to be taking it a step too far. Technology always has a chance that it can glitch or crash, and while a person can also make mistakes, I think that it is a lot easier to hold a person accountable for errors than a machine.
    Another issue I see is that in the event that the technology crashes, companies may not have workers with the proper skills to continue production if they rely to heavily on the robots. While the robots will take over some jobs, I think that a limited number of new jobs could come about by robot usage. There will probably need to be someone to check and make sure that the robot has taught itself to properly do the given task, along with a supervisor to overlook the work that is being completed by robots. If robots really grow in popularity, Google may need to hire more people to help create these robots.
    I also agree with Taylor in wondering how much these robots cost and how long it takes the machines to learn something. I am really curious to learn more about how the machines are able to teach themselves to do things. If it is by trial and error, this could be expensive for a company although they will end up saving money in the long run by not paying for labor. How long is a machine learning robot expected to work for? If robots are expensive and need to be updated/replaced every couple years then it may not be beneficial to a company to replace workers with robots.


Note: Only a member of this blog may post a comment.