Loop AI Advances Intelligent Automation Continuum Toward Deep Learning

HfS has stated repeatedly that we believe that the notion of Intelligent Automation is evolving along a continuum of overlapping yet complementary automation approaches. This continuum is advancing predominantly along three dimensions: Toward unstructured data, toward less well-defined processes and toward cognitive computing/artificial intelligence. Sitting down with executives of Silicon Valley-based startup Loop AI Labs provided yet more evidence underlining that assumption.

Loop AI’s main product is the Loop Cognitive Computing Platform, which provides fully unsupervised deep learning. Requiring no human supervision, the platform automatically interprets natural language and extracts a ready-to-use structure with concepts and relationships. It consists of proprietary algorithms and low-cost, custom-built High Performance Computing graphical processing unit (HPC GPU) accelerated hardware. The platform works on human capacity principles inspired by the neocortex, the brain’s center for language and reasoning. Loop AI’s proprietary algorithms learn the underlying structure and concepts within each organization’s dark data. Or, put in more simple terms, this approach allows companies to analyze up to 80% of their data that thus far has been unused. A key element in Loop AIs’s value proposition is to provide structures and semantic learning so that any other tool, for instance analytics or Intelligent Automation, can use these data sets.

While Loop AI is aiming to build out specific vertical solutions with system integration partners, some of its early use cases suggest broad applicability. For example, providing cognitive input and analytics for recommendation systems or anomaly detection on unstructured data for fraud detection and prevention. In addition, Loop AI Labs is working with automotive and insurance companies to understand the notes, statements, and jargon of auto mechanics, accident participants, and claims adjusters. By extracting structure from this unstructured data, Loop AI gives executives real-time visibility into what’s happening at the ground level, and a means to deploy intelligent automation based on that data. However, Loop AI is looking to go beyond these more process centric use cases by extending its deep learning approach to IoT scenarios. The thought process behind these scenarios is that current cloud models can’t support the bandwidth required for millions of devices to collect data. Therefore, Loop AI bypasses the internet by storing and processing data on the device, close to the source. The company has built prototypes and is working with mobile device manufactures (cell phones and wireless cards) and media companies (set-top boxes) to define proof of concepts.

Loop AI is adding a distinctive and complementary voice to the discussion of Intelligent Automation by extending scenarios around unstructured data to dark data. This underpins domain-specific learning that is unsupervised and language independent. By offering an HPC appliance that is optimized for its algorithms, Loop AI is not only overcoming performance issues because of the velocity and size of the data, but it enables deep learning models that are specific to a customer.

HfS will follow the progress of Loop AI Labs with great interest. Cognitive computing, machine learning and artificial intelligence will be key topics on our research agenda for Intelligent Automation in 2016 and we encourage other organizations with similar capabilities to share their experiences with us.