It’s been a huge year for a type of artificial intelligence known as deep learning.
Google, Twitter, and Yahoo have all acquired startups that deal in deep learning, which entails training systems called artificial neural networks on lots of information derived from audio, images, or other inputs, and then presenting the systems with new information and receiving inferences about it in response.
Researchers at big companies and academic institutions have been making inroads in deep learning in recent months. Facebook and Baidu have made major hires, while people have done research in the area at Netflix, Spotify, and other companies. Meanwhile, a new class of deep learning startups has cropped up, giving investors an area to watch closely.
Here are a few deep learning startups to pay attention to in the months to come.
This startup is determined to make an impact on the health care industry. Once its systems have trained on extensive libraries of medical images, Enlitic hopes to provide insight into new medical images that doctors might not necessarily be able to find on their own. The startup has raised a $2 million seed round.
Ersatz provides a general-purpose web service and on-premises software for training a neural network with words, text, and images and then using it to make predictions. The startup also offers hardware companies can buy and run in their own data centers. Ersatz launched in June.
During his time as a Ph.D. student at Stanford University, MetaMind cofounder and chief technology officer Richard Socher crossed deep learning with natural-language processing to develop a method he called a recursive neural network. Now the system is the basis of a startup with seed funding from Salesforce.com chief executive Marc Benioff, among others.
Nervana cofounder and chief executive Naveen Rao has observed the renewed interest in artificial neural networks, and early this year he left his job at chip maker Qualcomm to build deep learning technology, from a cloud software all the way down to specialized server processors. Nervana a $3.3 million round in August.
Skymind is the brainchild of Adam Gibson, who developed the open-source project DeepLearning4j. The idea of Skymind is to provide commercial support for companies that want to use DeepLearning4j and also incorporate deep learning into their Hadoop systems for storing, processing, and analyzing lots of different kinds of data. Skymind launched in June.
Deep learning talent is still in limited quantity, so don’t be surprised if some of these startups get swept up before the end of next year.