So much has happened this year in the world of deep learning, that trendy type of artificial intelligence that entails training artificial neural networks on large data sets and then getting them to make inferences about new data.
There have been technical breakthroughs, acquisitions, funding deals, open source releases in the field, and even the establishment of a nonprofit research lab backed by the likes of Elon Musk and Peter Thiel.
All of the startups I included in my roundup “5 deep learning startups to follow in 2015” (as well as others, like Clarifai) have made progress of some kind this year. Now, as we wrap up 2015 and get ready for 2016, a different set of deep learning startups are top of mind for me.
Here they are:
Here’s a startup that’s trying to use deep learning to get big in the antivirus software market. Deep Instinct officially launched last month. Chief technology officer Omid “Eli” David, who received a Ph.D. in computer science at Israel’s Bar-Ilan University, is a longtime computer chess researcher who published a paper (PDF) on deep belief networks for malware classification earlier this year. Deep Instinct’s investors include UST Global and Blumberg Capital.
Lunit, whose cofounders come from the DARPA Robotics Challenge-winning South Korean university KAIST, helps radiologists understand medical images. U.S. startup Enlitic also does that, but Lunit is different by virtue of its Data-driven Imaging Biomarker (DIB) algorithm and its focus on chest X-rays, cofounder and chief executive Anthony Paek told VentureBeat in an email. Earlier this year the startup participated in the ImageNet image recognition competition, going up against researchers from Google, Microsoft, Tencent, and other big technology companies. Last month Lunit raised $2 million from Formation 8, SoftBank Ventures Korea, and Daum Kakao’s K Cube Ventures.
This early-stage startup hails from the Swiss city of Lugano. The name Nnaisense “is pronounced like nascense, because it’s about the birth of a general purpose neural network-based artificial intelligence (NNAI),” cofounder and president Jürgen Schmidhuber told VentureBeat in an email. Schmidhuber is scientific director of the Swiss AI lab IDSIA, which is also located in Lugano. Schmidhuber has a track record in artificial intelligence and influence in the field as a result of his contributions.
TeraDeep has built technology for image recognition, which is a very common use of deep learning. Based in Santa Clara, California, the startup has a few YouTube videos demonstrating the applications of the technology. The startup offers Learning Camera software alongside application programming interfaces (APIs) that developers can incorporate into their applications. Facebook’s LeCun is an advisor of the startup, as is Clément Farabet, cofounder and chief technology officer of Madbits, the deep learning startup Twitter acquired in 2014.
In South Korea, Vuno has been busy forming partnerships with health care organizations to help doctors diagnose diseases with technology that makes inferences about medical data. The founders previously worked on speech recognition at Samsung. The startup uses a proprietary framework called VUNOnet but plans to open-source it, chief technology officer Kyuhwan Jung told VentureBeat in an email. Like Lunit, Vuno participated in ImageNet this year. Investors include Bon Angels and Future Play.