With CaffeOnSpark, Yahoo joins the open source AI fray


By Michelle Goodman

The race to the front of the machine learning pack continues. This week Yahoo open sourced the code to its CaffeOnSpark artificial intelligence engine for researchers, developers and competitors to use and modify as they wish.

Yahoo joins Google, Microsoft and Facebook in open sourcing their machine learning frameworks to the worldwide computing community. The hope, of course, is that the AI contributions of developers, researchers, startups and bigger companies around the globe could help propel any one of these tech titans to the head of the machine learning class.

CaffeOnSpark's distinction is that it employ deep learning, a still relatively mysterious type of artificial intelligence that allows machines to learn from vast collections of online data. For instance, engineers can use deep learning to program computers to recognize speech patterns or identify the contents of photos or videos.

Yahoo's internal teams have used CaffeOnSpark on a handful of projects during the past few quarters, according to Hadoop, a Yahoo Developer Network blog. Most notably, Flickr engineers used CaffeOnSpark to greatly improve the accuracy of recognizing images in millions of photos. Much of the internal feedback about CaffeOnSpark has been positive, the blog said.

It's not Yahoo's first foray into sharing in the AI world. Just last month, the company released what it called the largest-ever machine learning dataset to the research and scientific community.

For more:
- read the Yahoo blog post

Related Articles:
Yahoo brings machine learning data to the research community with huge dataset release
Google continues to push the AI envelope with release of new toolset that lets robots 'see'