At the beginning of the year, Google CEO Sundar Pichai discussed the company’s key areas of focus. He said that, aside from allocating some funds for investments in AR, VR, and cloud technologies, Google will be an “AI first” company. This statement could very well explain one of Google’s latest takeovers.
Google has acquired Belarusian startup AIMatter. As usual, the terms were left undisclosed. The company was originally funded by Haxus, an investor that has already backed a number of AI projects, including the Russian-built photo filtering app Prisma.
AIMatter first drew attention with Fabby and Fabby Look, advanced image processing apps that use neural networks to transform images and videos in real time. Google’s acquisition of the company shows the search giant’s belief in the startup’s potential to transform image processing.
Neural networks in your hand
According to a Haxus report published soon after news of the acquisition broke, the AIMatter team came up with a next-generation image processing technology that allows it to “run deep learning models in realtime directly on smartphones.” In other words, it puts something as complex as computer vision into a lightweight, offline app.
Prisma Labs was one of the pioneers in AI-based image processing technology that’s fast enough for the mobile environment. Soon after debuting a location-based feed, Prisma introduced a new product for the web that allows users to create their own filters.
In turn, AIMatter introduced a way to process portrait images and video to recolor, refine, and intelligently extract and transform parts of visual content. The computer vision side of this technology is based on layers of neural networks trained on servers that run directly on smartphones. This makes the solution faster and more elegant than its predecessors. However, the true prize exists far beyond the current application.
Why this innovation matters
AIMatter’s SDK for image processing is a unified AI-powered toolkit with rich potential. In fact, this technology already has a high potential for application.
First, it showcases the use of computer vision in everyday life. The application of self-learning neural networks allows machines to detect and respond to certain visual cues — something that is easy for a human brain to process, but complex in terms of math. Eventually, this type of technology will free people of monotonous tasks and allow them to take on more intellectually complex responsibilities. This branch of AI is practical and seen as attractive, especially in process automation and driverless technology.
Second, advanced image processing capabilities contribute to visual search development. The broad application of this technology grabbed the spotlight thanks to products from Google, Pinterest, and eBay. eBay’s ShopBot on Messenger already allows users to send a photo of goods — say, sneakers or glasses — and search for similar items in a store.
Finally, advanced image processing that is lightweight enough to use on a personal device will become accessible with image processing innovation. The technology could be available to all smartphone users as a way of complementing the mobile UX that already goes beyond a graphic interface to become a hybrid of text, voice, and image.
It’s not just about the tech
With Google’s strong backing, the technology introduced by companies like AIMatter can be applied in a variety of unpredictable ways. The same goes for the potential of the talented individuals who built the technology.
Giants like Google keep their attention on the startups that grow in the AI ecosystem, be it computer vision, robotics, or NLP-powered chatbots. In fact, this is the best way for monoliths to follow the AI first approach. More than half of companies engaged in AI have under 50 employees and hold the strongest asset in the industry — talent.
Katherine Lazarevich is cofounder and managing partner at Digiteum, a digital technology agency.