Machine Intelligence, Friend or Foe?

SICS Swedish ICT, in collaboration with strategic partners, is establishing a new five year research center, Learning Machines@SICS, for advanced research and applications in Big Data Analytics and Machine Learning. The research center will focus on high value, high impact applications of machine learning on massive data sets.

Recent years have witnessed an unprecedented shopping spree among the global Internet giants for machine learning expertise. World leading researchers such as Geoffrey Hinton, Yann LeCun, and Andrew Ng have been snapped up by Google, Facebook, and Baidu. In one famous example in 2014, Google acquired the fledgeling UK machine learning company DeepMind for a whopping 5 billion SEK, Google's largest European purchase, to secure the expertise of their around one dozen researchers.

So why is Machine Learning suddenly so important and valuable? The recent events can be traced back to a technical breakthrough in 2006, when Geoffrey Hinton and others mastered learning for ”deep neural networks”, which are able to capture the characteristics of complex data sets in multi-level hierarchies of concepts with millions and even billions of parameters. As one example of the impact, in 2012, a deep neural network dramatically outperformed the competition in the ImageNet visual recognition contest, beating decades worth of human efforts on carefully hand-crafted machine vision algorithms.These methods are widely applicable in many domains, and are believed to herald a fundamental transformation of many engineering disciplines, where machine learning approaches will outperform and make irrelevant human designs.

Machine learning will necessarily replace human labor, as have other technologies before, and will similarly transform industry and society. With the greatly increased capabilities of intelligent machines, we have also seen a surge of warnings on the future existential risks of Artificial Intelligence from Stephen Hawking, Elon Musk, and many others. We are deeply optimistic about the potential of this technology to transform industry and society for the better, and the risks can only be handled by mastering it. We believe that Sweden must lead and be among the first to benefit from this transformation.

SICS has been driving Machine Learning, Neural Networks, and other methods for statistical modeling of complex data since the early 1990s. We are uniquely poised to leverage our expertise, the recent developments in the field, and the interests of our partners in creating new valuable applications for Swedish industry and society in the new Learning Machines@SICS center.

View all stories

 Previous  Swedish ICT Annual Report 2014 Next