Roberto Baldwin was a reporter for The Next Web in San Francisco between April 2014 and March 2015. Roberto Baldwin was a reporter for The Next Web in San Francisco between April 2014 and March 2015.
Deep learning, machine learning and neural networks sound like the features of a pulp science fiction, but they’re real and the technology is now available to any company without having to hire data scientists to build your own system from scratch.
Ersatz Labs has launched a deep learning platform that’s available to anyone that wants this type of analysis. It’s like the Amazon Web Services (AWS) of machine learning. The service is has a web interface and API. Both can be used if a company needs to determine, for example, how stocks will behave based on a wide range of data, it can do so without having to build its own system.
That doesn’t mean that this is a turn-key solution for anyone wanting to analyze data. The system still needs to be “trained” with your data and what you expect out of that data which means you have to know what you’re doing. But, by offering an interface layer on top of a set of powerful tools, its cutting down on the work needed to spin up a system like this. “We didn’t invent deep learning by any means, but we’re trying to make it easy,” says Ersatz CEO, Dave Sullivan.
It’s still a pretty tough subject to wrap your head around. Deep learning divides data into several networked stacks to to learn about that the information in the data. Once the information has been collected machine learning actually learns over time how those large chunks of data relate to one another while forming connections.
Ersatz can be used in the cloud or on hardware installed on site. Both solutions use specialized GPUs that can eat through data 40 times quicker than CPUs. The cloud solution charges per-minute while training the system with your data. The hardware solution offers unlimited data crunching with a specialized appliance.
Get the TNW newsletter
Get the most important tech news in your inbox each week.