As we enter the twenty-first century, mankind has gotten closer than ever to producing artificial intelligence, or in simpler terms, machines that are able to think in the same manner that humans do. Deep learning, a form of AI, is essentially data being processed through numerous layers of artificial neural networks.

How It Works:

The “deep” in the term refers to the depth of the neuron network, in which large amounts of data passes through. The different layers’ neurons would be created to expand upon the work of the one before it, so basically information would be fed into the first layer, which would take out basic observations. Then, it is passed on to become the input that the next layer receives. This second layer would then extract more data, which progresses from layer to layer, on and on until the system solves the recognition task.

In order to produce more accurate results, the deep learning systems are fed extremely large samples of data in order to account for as many variations as possible. For example, in the case of recognizing handwriting, by giving the system a large sample, it would be able to learn more about handwriting and produce more accurate results.

Google self driving car, free image, creative commons

Google’s self driving car, an example of artificial intelligence technology that will be widespread in the future.

Practical Uses:

Deep learning can be utilized with virtually any problem set, either to find a predetermined goal, like specific classification, or to find hidden patterns in unstructured data.

It has the ability to process enormous amounts of data from the Internet, carry out facial and voice recognition, and control self-driving cars, in addition to other endless possibilities. One of the most revolutionary uses of deep learning would be for medical science. The system allows the machines to scan information exponentially faster than humans could, narrowing out different variables in the science and medical field, even in molecular activity.

In one of MIT’s latest projects, the engineers designed what is deemed the “Nightmare Machine”, which utilizes deep learning to study what kinds of imagery elicits fear in humans. After examining over 200,000 images, the system can now alter average images into scary ones. Although this may seem an insignificant step, it is principally teaching AI to better comprehend human emotions. You can see some examples of these modified images at nightmare.mit.edu.

To show you how the process works, here is a normal picture of the Taj Majal:

Taj Majal, source: nightmare.mit.edu

nightmare.mit.edu

After running it through the “Nightmare Machine” using deep learning:

taj mahal, source: nightmare.mit.edu

nightmare.mit.edu

As technology and science continues to progress, deep learning and auto intelligence will pave the way for the advancement of human society.

Previous post

Two Electric Giants: Tesla and SolarCity

Next post

The War on Malaria: Genetic Engineering

Sophia Xiang

Vice President

No Comment

Leave a reply

Your email address will not be published. Required fields are marked *