How do neural networks think??

Artificial Intelligence is slowly taking over the world with its abilities.They are able to do this with the help of machine learning systems called neural networks.Understanding the working of these systems can help in transforming the way artificial intelligence works.Moreover, we will be able to improve them and build better machines.

The researchers from MIT computer and Artificial Intelligence Laboratory have shown a new way to know neural network that applies to the system that can text as input and produces a string of symbols as output.They believe that this technique can be used as a form of translator for online natural language processing systems.In fact, this technique can also be used for interpreting black box -text processing systems.

The Test

The researchers applied this technique to 3 different types of natural language processing systems.The first one was a simple pronunciation while the second one was a set of translators, one human, and the other one was a computer.The third was a simple computer system dialogue.
The analysis of the translation systems showed high dependencies between individual words.
Ironically the systems like object identifying systems commit errors.If a picture of the dog is shown, then the system may identify it as 70% dog and 30& cat due to the similarities.Similarly, this technique has limitations too.This sentence compressing network supplies alternatives for
each word in decoded systems along with the probability of the sentence being incorrect.

Comments

Popular posts from this blog

5 Ways to become smarter everyday

How can foot ulcers in diabetic patients be detected??