What you missed in Big Data: The AI arms race continues

man machine brain gears magnifying glass ai artificial intelligence thought

Three months since Google Inc. fired the first shot, the one-upmanship in the data science ecosystem shows no signs of slowing down. Last week saw Yahoo Inc. return to the fray after a brief hiatus and open-source a homegrown extension for Spark that extends its machine learning capabilities into the realm of artificial intelligence.

CaffeOnSpark enables organizations to exploit the data crunching engine’s processing power to increase the performance of their neural networks, which are much more complicated than the garden-variety algorithms currently used in most analytics projects. Whereas a traditional machine learning model is hard-wired for a specific purpose, like identifying the presence of a certain object in photos, Yahoo’s extension makes it possible to have an image recognition service actively try to find new kinds of objects that might be of interest to its users.

But while CaffeOnSpark may simplify the implementation of such algorithms, creating neural networks is still beyond the ability of most organizations. The task not only requires highly specialized talent that is neither easy nor cheap to come by but also expensive servers packed with graphical processors. Facebook Inc.’s artificial intelligence set out to tackle the latter challenge the day after Yahoo released its extension by launching a new program that aims to give data scientists access to the kind of hardware necessary to develop complex models. In the first phase of the effort, the social networking giant will donate 25 state-of-the-art GPU servers to several research institutions in the European Union.

The venture capital community, meanwhile, is also starting to recognize the growing importance of machine learning. A group of investors led by Tenaya Capital last week poured $35 million into the coffers of ThousandEyes Inc., a monitoring provider that employs automated algorithms to pinpoint traffic bottlenecks in corporate networks. Its service can keep track of both an organization’s internal switches and routers as well as the carrier infrastructure that connects its infrastructure to the outside world.

photo credit: geralt