Google co-founder Sergey Brin believes the best search engine would be like the HAL 9000.

The next step in artificial intelligence is here, and Google thinks that it's worth investing in.

Computational power is smaller, faster, and cheaper than ever before. Object recognition and real time speech translation are reaching impressive levels of completion.

Ray Kurzweil is a famous inventor, author, futurist, and most recently Google's new director of engineering. Why would a fabulously rich and successful inventor and entrepreneur go to work for a company when he has always worked as the head of his own companies? The answer is simple: Unrestrained use of Google's collective processing power.

“This is the culmination of literally 50 years of my focus on artificial intelligence,” - Ray Kurzweil

The software learns to recognize patterns in digital media like music, sounds, images, and video. Showing media to software to allow it to gradually build a consciousness is a decades old idea- after all humans learn the same way. In our contemporary digital age mathematical formulas and computing power have increased tremendously from the 70's, when the first machine neural networks were built.

In recent years Google has become a specialist in the previously obscure field of deep learning and artificial intelligence in general. It makes a lot of sense when you think about it- the world's most popular search engine is being developed to better understand what we want and how we are saying it.

Applications?

Google has used the combined parallel processing power of 16,000 processors in an image recognition experiment. The simulated neural network had over a billion connections. Images from 10 million randomly selected images were shown to the system. Objects and themes were correctly placed in a range of 22,000 categories 16% of the time. This is a 70% improvement from contemporary alternate methods of image recognition. When the number of categories were reduced to 1,000 the system sorts the images correctly over 50% of the time.
The current version of Android's voice search uses deep learning. Its errors were reduced by 25% when this technique was applied, and many technology critics now consider it to be even better than Apples Siri.
Youtube image search could be improved.
Self-driving cars could see the world better using this technology.
Machine vision e.g. Industrial inspection and robot guidance.

It should be noted that this new technique has its critics. Jeff Hawkins, the founder of Palm Computing claims that true intelligence stems from the advancement of motion retention and recognition, and not huge piles of images like Google uses. For example when you watch a dog running, you learn to recognize animals from the way they move- we don't remember a series of still images like those Google used in its experiment.