How will Machine Learning Change our lives?

The honest answer to this question and whether Machine Learning will take over the world is extremely complicated. I believe the honest answer is yes and no(as it is a good answer to a lot of questions in life).

Machine Learning in its most basic form is theoretically only as smart as the person programming the machine, and the strength of the learning algorithm. If we just use logistic regression we can get a machine to behave about as accurately as a human would. This will continue to affect the job market by reducing the need for certain types of unskilled workers. In fact, the classic machine learning problem of reading handwriting has shown that a machine can classify numbers with 98% accuracy(about as well as a human)[1].

Machine Learning however suffers from the fact that it cannot define a problem because that in itself is such a subjective thing. Since there are at least 10¹⁰ unique problems it seems strange to think that a machine could be trained to recognize them. Sure, the machine can create its own groups, but what if we don’t know we have a problem? Can innovation be taught? I believe that is not possible because that requires empathy. The deep empathy that most of us possess is something that I do not believe you cannot teach a machine. Sure, you can apply reenforcement to add something that will call a machine to action when someone is hurting, but that is not empathy. Empathy is a choice, and without that choice, it is a duty.

Of course, Machine Learning is very powerful. There are algorithms that can analyze mammograms very well(94% accuracy in 2010!)[2]. Standard approaches have an accuracy of 84%! Google self-driving cars have been about as good at driving as their human counterparts, so it is undeniable that Machine Learning is incredibly powerful, and it is possible that machines can do things better than we can. But this isn’t a new phenomenon. Machines have been better at sowing, but we still pay for designer clothes because they are imperfect. Automobiles Robots still need humans to check the quality of the car.

The strength of machine learning is in its capacity to learn complex algorithms with regards many parameters. A machine as structured now is really just an adaptive function that learns its own rules. Machines do not have the capability to function at higher levels of thought and abstraction. I believe we should be safe from sentient terminators, but it is possible more and more jobs will become automated and that machine learning will become present in almost every facet of our lives.

Sources:

  1. http://neuralnetworksanddeeplearning.com/chap1.html
  2. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4220564/