Deep learning

For a long time , if you wanted to make a computer do things , you had to specify every single step of its movements when you wrote the code. Actually, this thing still happens , but a particular algorithm will be able to simplify and improve this field. It is called Deep Learning.
A first example that shows the potentiality of this method is figured by a challenge accepted by Arthur Samuel. He wanted to get its computer to be able to beat him at checkers. His idea was to make computer play with itself and learn independently new strategies for winning. Surprisingly, it worked ! It had beaten a checkers champion! ( in 1962 )

Nowadays there are companies that are studying this incredible technology, such as Kaggle , which is trying to solve previously unsolved problems.

Perhaps the first big success of machine learning was the Google search engine, which can find information for us. Also Amazon or Netflix is usually able to suggest the right product .

Another example is an IBM machine which was able to answer complex questions in a TV game ("Jeopardy") or even the self driving cars which are becoming better and better.

However these algorithms have learned how to do this from data rather than being programmed by hands.

The amazing thing about these algorithms is that they can perform tasks better than us. They can even do things we don't know how to do ourselves. For instance , in the University of Toronto , a guy ( Geoffrey Histon) led a team in a competition, and they discovered drugs automatically, without almost any background in chemistry or biology.

This is inspired by how the human brain works . The more data and computation time, the better it gets .There are non theoretical limits .

You can even talk in English, and in real time , the machine will tell the same thing in Chinese , with the same voice ; and all of this was done by deep learning.

The idea behind Deep Learning is not to replaced humans by a computer, but to make them work together.

We're going to replace something that used to take a team of five or six people about seven years and replacing it with something that takes 15 minutes for one person acting alone.

This process is very fast and it takes four or five iterations. Humans can check if the deep learning algorithm has made mistakes and if we find one, we can let the computer know about it

Made with Adobe Slate

Make your words and images move.

Get Slate

Report Abuse

If you feel that this video content violates the Adobe Terms of Use, you may report this content by filling out this quick form.

To report a Copyright Violation, please follow Section 17 in the Terms of Use.