Google’s DeepMind Uses News Articles To Teach Computers To Read

The Daily Mail reported this story, making it perhaps the ultimate symbol of the recursive nature of deep machine learning.

(June 18, 2015) – London, UK 

Google acquired the artificial intelligence company DeepMind in January, 2014. Dr. Demis Hassabis is the co-founder and CEO of the London based machine learning company which aims to build true artificial intelligence with the mission to solve some of the most complex and pressing issues confronting our society.

As reported by Danny Palmer (Computing, UK):”We might have to come to the sobering realisation that even with the smartest set of humans on the planet working on these problems, these systems may be so complex, that it’s difficult for individual humans and scientific experts to have the time they need in their lifetimes to even innovate and advance,” said Hassabis.”It’s my belief we’re going to need some assistance and I think AI is the solution to that…The reason I work on this, rather than working on any of these specific scientific problems, is I see AG [artificial general intelligence] as a solution to all of these problems,” he said.

There has been a revolution in deep learning because of two main breakthroughs in the field: (1)  Computer scientists have developed methods to train networks with many layers,  called “neural networks” which are primed to learn. (2) Learning algorithms can be developed for massive data sets which look at grammatical links and causal relationships between words.

What is the current breakthrough approach announced by DeepMind?  The structure of many news stories online includes bullet points containing main story points independent of the body of the story. According to MIT Technology Review, DeepMind has created a database of over 328,000 news stories by the Daily Mail website, MailOnline, and CNN online which have this structure and could be utilized to both teach artificial intelligence and test it.

The Daily Mail reported this story, making it perhaps the ultimate symbol of the recursive nature of deep machine learning.

 

Video: Embedded via YouTube/ZeitgeistMinds (https://youtu.be/2s4D-8MpreE)

 Related Media:

 

Subscribe@Raw Science TV |Facebook@Raw Science |Twitter@RawScienceTV |YouTube@RawScience1

 

Photo credit: Chris Isherwood (Creative Commons)

Be first to comment