Machine learning development is accelerating at an unprecedented pace as new breakthroughs and applications are being announced or demonstrated on an almost weekly basis.
Driving the acceleration is a combination of advances in machine learning methods, such as deep learning, advances in hardware and GPU-accelerated computing, as well as a continuously increasing focus in machine learning R&D by all the major technology centers, universities, and giants such as Google, Facebook, Microsoft, Baidu, and the like. Further, development is now in a positive feedback loop, thanks to the extraordinary openness of the AI community, demonstrated both by the pace of results publication as well as the adoption and contribution to open source initiatives, by academia and industry alike.
It is no surprise then that open source frameworks and tools have become the de facto standard and best available technology in this field. Microsoft, in particular, has proven to be a champion of this transformation, topping the list of organizations with the most open source contributions on GitHub.
The Difference Between Horizontal and Vertical AI Start-ups
The evolving venture capital landscape is another indicator of the acceleration of machine learning and AI in general. While it is a fact that venture capital funding is rushing into machine learning and AI startups, growing according to CB Insights from $589M in 2012 to over $5B in 2016, it’s important to distinguish between horizontal AI startups delivering generic tools and AI services across industries, and vertical AI startups that solve full-stack industry problems, as correctly pointed out by Bradford Cross, founder and partner of DCVC, the world’s leading machine learning and big data venture capital fund. We fully share his view that if a company isn’t solving a full-stack problem, it will soon face an increasingly commoditized world of “shallow” AI services, and will likely end up either being acquired or wound-down due to lack of traction.
Vertical AI startups, such as eSmart Systems, solve full-stack, concrete industry problems that require deep domain expertise, access to unique proprietary data, and a product that utilizes machine learning and AI to deliver its core value proposition. This puts eSmart Systems in the unique position of being able to identify real customer needs at all levels of the organization that can be better met with AI, or new needs that can’t be met without AI.
Read also: Using Machine Learning and Smart Meter Data for Fraud Detection
Artificial Intelligence to Support Power Utilities
Among other applications, eSmart Systems focuses on using AI to meet new load management needs of utilities, as well as their increasing needs for intelligent asset management. This is achieved, for example, by advanced load prediction models at different aggregation levels, predictions for distributed generation and storage, and with deep learning enabled drone-based inspection of power transmission and distribution infrastructure. These AI implementations are at the core of two of eSmart Systems’s full-stack solutions, Connected Grid and Connected Drone respectively, that support power utilities in their day to day operations and deliver tangible value to our customers.
Read also: Why AI and Machine Learning are Essential for Next Gen Utilities
Savings in Existing Infrastructure
Taking the example of Connected Grid as a case in point, the concrete feedback gathered from two early adopter mid-sized Norwegian DSOs reveal that in one year, these two DSOs combined saved about $1.5M in faster fault detection and recovery, and close to $4.7M in smarter investments and better use of existing infrastructure. A conservative estimate of $6M savings per year for these two DSOs extrapolated to the whole country, shows potential savings of about $240M per year; This with existing infrastructure and technology in Norway alone.
Another example is Jacksonville Electricity Authority (JEA) and how they used advanced drone technology and machine learning to restore power in the aftermath of Hurricane Irma in 2017.
These references are examples of how machine learning and AI technology will continue accelerating and I am convinced that start-ups with full-stack vertical approach will continue strengthening strategic position and market share.