IBM, Nvidia and Amazon have each announced new artificial intelligence tools and services for data scientists and developers. The ecosystem surrounding deep learning (talent, tools, infrastructure), while holding potential, remains underdeveloped, as is its ability to scale.
Nir Bar-Lev, co-founder and CEO of Allegro.AI, sees a huge opportunity for deep learning computer vision. This type of technology can be used in autonomous cars, drone navigation systems, medical devices, surveillance cameras and many other devices that need to process and learn huge datasets to make immediate (even life-saving) decisions.
Digital Journal caught up with Nir Bar-Lev to learn more about the technology.
Digital Journal: What are the advantages of deep learning for businesses?
Nir Bar-Lev: Businesses are exposed to a vast ocean of data, which continues to grow exponentially every day. Using deep learning – a subset of machine learning, computers learn to classify patterns, and understand information that helps organizations make better decisions that are intrinsic to their growth and success.
Like machine learning, deep learning allows businesses to analyze huge datasets and arrive at insights that are simply not feasible with traditional human-based analysis methods. However, deep learning goes one step further because it builds models that can tackle even more complex problems.
It is used primarily in computer vision, natural language processing and speech recognition where the problem of building a “model” to comprehend a given image or understand a piece of text or spoken language is so complex, that human models have hit a performance ceiling.
Using deep learning, for example, we have been able to bring speech recognition to over 95% accuracy levels, and have been able to use face recognition to identify people more accurately than with the human eye. While machine learning usage is currently more prevalent than deep learning, we can expect to see deep learning techniques increasingly replace machine learning in the future.
DJ: What do the IBM, Nvidia and Amazon new AI tools and services for data scientists and developers signal?
Bar-Lev: This year, we’ve seen tech titans including IBM, Amazon, Microsoft and others announce plans to develop tools for developers and data scientists that remove the barriers of training deep models. Global IT analyst firm Gartner predicts that by the end of the year, a whopping 80 percent of data scientists will have deep learning in their toolkits – so it’s no surprise that every major technology company is investing heavily in this space, and scrutinizing how they can deliver business value to their customers.
DJ: To what extent is the ecosystem around deep learning — talent, tools, infrastructure — underdeveloped?
Bar-Lev: When we founded Allegro.AI, my two co-founders and I recognized that deep learning breakthroughs reside in research and academia, and as such, a significant opportunity exists to develop an entire ecosystem to help companies scale their AI solutions.
AI is a very intensive technology, and the real benefits come from its ability to process huge datasets to propel solutions in real-time – whether it’s robots, self-driving cars, surveillance cameras, or other applications where the processing power should be done on the device. But because building and developing for AI solutions is intrinsically different compared to other business applications. For example, there’s a conspicuous dearth not only of data scientists, but also of PhD research scientists that can effectively counsel companies on deep learning implementations.
While basic tools do exist, there’s a severe lack of infrastructure to effectively develop AI products at scale. Deep learning is still a nascent technology, and many companies are still trying to figure out their strategies when it comes to taking full advantage of what it can offer. As organizations start to design and develop new solutions, the talent, tools, and infrastructure will start to advance over time.
DJ: What is deep learning computer vision?
Bar-Lev: Deep learning computer vision merges two important and innovative technologies today. Computer vision is a part of computer science that teaches machines to understand and identify video or images, while deep learning is a subset of machine learning, which itself is a subset of AI.
By combining both technologies, companies can better analyze, process and understand digital images and video in real-time. This is critical for applications such as autonomous cars, for example, which need this to drive along busy streets and highways to avoid crashes from occurring, or for drone navigation systems that similarly need to make rapid decisions to best navigate around their surroundings.
DJ: Which sectors will benefit most from deep learning computer vision?
Bar-Lev: Any industry that needs to process vast amounts of images or video to make instantaneous decisions will benefit most from deep learning computer vision. We’re working with a number of organizations that are looking at applications in transportation, healthcare, construction, and many others.
DJ: Can you provide some practical examples?
Bar-Lev: One clear example of how deep learning computer vision is being used is in autonomous cars. Imagine being in a self-driving car that’s cruising down a highway, but needs to brake suddenly to avoid a crash. Deep learning computer vision can quickly understand its surroundings to prevent a collision from happening. A drone navigation system is a similar example, as it needs to “see” what’s happening as it navigates from point A to point B.
Another excellent example is in diagnostic applications in the healthcare industry, which can rely on data sources – such as rich forms of medical imagery – to accelerate the AI diagnostic process. While diagnosis is very complex, deep learning computer vision can assist physicians in making sense of live health data to support treatment.
These are just a few examples, but we’re just starting to scratch the surface of what deep learning computer vision can deliver. Organizations are only beginning to utilize AI’s full potential for intelligent devices today, and we can expect to see many more applications to surface in the near future.
DJ: What services does Allegro.AI offer?
Bar-Lev: Allegro.AI is a deep learning computer vision platform that enables organizations to deliver the next wave of AI-powered technologies. We are the first end-to-end AI lifecycle management solution, which simplifies the process of developing and managing AI-powered solutions, such as autonomous vehicles, logistics systems, drones, and many others. Our customers can bypass the manual and time-intensive exercise of building their own datasets and models by using the Allegro.AI platform, and allow them to keep complete and confidential control of their data and AI assets.
DJ: How did you seek funding?
Bar-Lev: We had not planned start a Series A funding round so soon, but we had been approached by several investors – including our lead investor – who saw the potential and opportunity of the Allegro.AI platform. Our lead investor, MizMaa Ventures, is highly-regarded as a venture capital firm that works with deep tech innovators emerging from Israel, and we felt that they were perfectly positioned to launch the company and expand our footprint to the U.S. and Asian markets.
Our other investors – Robert Bosch Venture Capital, Samsung Catalyst Fund and Dynamic Loop Capital – are all very well-respected. They will be instrumental in helping us develop our platform, and build and open ecosystem that empowers companies to leverage the power of AI.
DJ: What are your goals and long-term vision now that you’ve launched?
Bar-Lev: In the short term, the $11 million cash infusion we’ve received will help us continue to develop our platform, roll out our solution to more customers, and hire top talent. We see a huge opportunity to fill the gap in providing an underlying platform to deliver the next wave of intelligent devices in virtually every industry — from transportation, healthcare, smart cities and many others. Our long-term vision is to be the underlying infrastructure that really delivers the next wave of intelligent devices. Simplifying the AI development process is a huge undertaking with an even bigger opportunity, and we’re looking forward to it.
