Saturday 26 July 2014

What Breed does that Dog Belong to? – Let Microsoft Answer that!


Project Adam
Microsoft’s research division is very proactive and always making headways into cutting-edge research and pushing the boundaries of software and hardware. One such innovation is Microsoft’s project Adam – an advanced AI with a manifold performance increment over its predecessors and its competitors along with advanced object recognition technology it incorporates.

Project Adam

At Microsoft’s 5th Research Faculty Summit keynote, Harry Shum, executive vice president of Technology and Research, demonstrated the power of the AI along with its various capabilities. Its primary goal was to recognize any object given to it. It was achieved by culling a massive 14 million image database from Flickr made up of 22k different user-generated tag categories. Using 30 times fewer machines than average, it was used to train Adam – a 2 billion connection neural network and it produces results 50 times faster than the competition.

Adam’s Capabilities

As actions speak louder than words, so a prompt demonstration was given and dogs definitely steal the show. In main aim was to recognize the dog’s breed. In order to make it more interactive it was integrated with Cortana, Microsoft’s very own personal assistant for their windows phone 8.1 lineups. Project Adam researcher Johnson Apacible pointed the phone’s camera at a Dalmatian named Cowboy and it was quick to identify the breed. He then showed a Rhodesian Ridgeback and it was spot-on to identify it too. A Cobberdog lead to some confusion as people said it was Labradoodle whereas Cortana said terrier but both were right. Just for some fun, he then pointed at Shum and Cortana was quick to reply with a negative.

The Possibilities

At the end, Shum quoted that there was a paradigm shift from personal computing to a future where user is the centre and hence it’s not about computing power or storage but people’s time and attention. One day Project Adam might make it possible for people to snap pictures of their skin or infection, identify the cause, and take appropriate medication based on the suggestions. It can also identify the calories contained in a meal by pointing it at the food. Another scenario would be when you are in woods; you would like to differentiate between poisonous and edible plants where it can help too.

Competition

However, as always the case, Microsoft is not alone in this race. In January, Google outbid Facebook to buy “DeepMind” – a London based AI Lab for 400 million. Further Google launched Quantum Artificial Intelligence lab to further study how quantum computing can lead to advances in artificial intelligence. HartmutNeven, Director of engineering for Research team at Google, was quoted saying that machine learning will gain utmost importance in near future with more accurate predictions and better models, which will also help them to produce a more useful search engine.

The Future 

In future, we might see Microsoft integrating it with smartphones and Cortana for a richer user experience and make them stand apart from the competition where product differentiation is becoming increasingly difficult.

No comments:

Post a Comment

Note: only a member of this blog may post a comment.