Our new Disruption by Blockchain series aims to highlight companies that are leveraging the incredible potential of blockchain technology to disrupt and revolutionize their industry. Through one on one interviews, we'll speak directly with industry leaders to cut beyond the hype and get directly to the heart of practical use cases and examples of how it will change the world, one industry at a time.
The following is an interview we recently had with Maxim Prasolov, CEO of Neuromation.
1. What’s the history of Neuromation? How and where did you begin?
MP: Neuromation began as an AI startup. As we began to look for ways to tap into computing power for training machine learning algorithms, we realized that blockchain mining farms were a perfect candidate. We created the Neuromation platform because we believe it will save our clients (and other companies interested in AI) money and time in developing machine learning algorithms. As demand for AI grows, Neuromation’s platform will be well-positioned to provide synthetic data and computing power to mid-size companies that don’t have the budget to build out infrastructure themselves.
2. Who are the founders and key team members?
Maxim Prasolov, CEO
As a multimedia producer during 2001 – 2017 Maxim has produced more than 50 animated commercials, 3D-movies, commercial and industrial applications, computer games. Worked with international retail and industrial brands such as Unilever, Yukos, TPE, Metro Cash&Carry, Severstal Group, Ferrexpo, commercial banks, investment and insurance companies. Prasolov has been a member of the IPO team of Ferrexpo on the London Stock Exchange, the biggest iron ore producer in Ukraine. Since 2014 he has been investing in drones development, AI, AR multimedia start-ups.
Constantine Goltsev, Chairman
Serial entrepreneur and online advertising industry veteran, Constantine has more than 20 years of experience in software and product development. He is the former CEO and founder of the pioneering video advertising network AdoTube, which grew from humble beginnings to 200 employees selling in 23 markets with 13 offices worldwide. AdoTube subsequently was sold to Exponential Interactive. He is a Founder and President of SolidOpinion — a company that has revolutionized engagement in online communities around content.
Yuri Kundin, Chief Operating Officer
Yuri Kundin is the former Director at KPMG US, San Francisco office.
Yuri has over 15 years of experience in advising both startups and large corporations in regulatory compliance matters. He has an extensive experience working with companies to address requirements of various agencies in the US including SEC, OCC, FDIC and international MAS (Monetary Authority of Singapore), HKMA (Hong Kong Monetary Authority), Russian Central Bank.
In the past 3 years Yuri has been specializing in developing framework and methodology for risk, compliance and attestation in blockchain, cryptocurrency and ICO eco-systems to help new emerging technology clients with innovative compliance strategies. Yuri holds a bachelor degree in public accounting and minor in information systems from Pace University, New York. He is a key member of the global KPMG community service team and has helped organize beneficiary events in orphanages, elderly care-homes and hospitals around the world.
Sergey Nikolenko, Chief Research Officer
Researcher in the field of machine learning (deep learning, Bayesian methods, natural language processing and more) and analysis of algorithms (network algorithms, competitive analysis), Sergey has authored more than 120 research papers, several books, courses “Machine learning”, “Deep learning”, and others. Extensive experience with industrial projects (Neuromation, SolidOpinion, Surfingbird, Deloitte Analytics Institute).
David Orban, Adviser
David is the Founder and Managing Partner of Network Society Ventures, a seed stage global investment firm focused on innovative ventures at the intersection of exponential technologies and decentralized networks. He is an entrepreneur, visionary, guru, author, blogger, keynote speaker, and thought leader of the global technology landscape. His entrepreneurial accomplishments span several companies founded and grown over more than twenty years.
David is the Founder and a Trustee of Network Society Research, a London-based global think tank. It is creating a vision and analytical tools to allow individuals, enterprises and the society at large to deal positively with the unstoppable transformation to a world based on decentralized exponential technologies that are disrupting the traditional centralized and hierarchical functions of governments and corporations.
Andrew Rabinovich, Adviser
World leading scientist in Deep Learning and Computer Vision research. Andrew has been studying machine learning with an emphasis on computer vision for over 15 years. He is the author of numerous patents and peer-reviewed publications. Andrew founded a biotechnology startup that was acquired. After receiving a PhD in Computer Science from the University of California, San Diego in 2008, he worked for years on leading R&D projects at Google. Currently, Andrew is a Director of Deep Learning at Magic Leap.
3. What problem are you solving? Who are you solving it for?
MP: Neuromation makes it faster, cheaper, and easier to train machine learning algorithms. Our goal is to speed up AI adoption for all kinds of companies.
Currently, it’s time-consuming and expensive to curate labeled datasets for training machine learning. The same goes for building the hardware infrastructure to train those algorithms. As a result, it’s difficult for small and mid-sized companies to develop custom AI applications.
4. What is your solution to this problem?
MP: We solve these problems by creating synthetic data for use in training machine learning algorithms. Then, we make it profitable for cryptocurrency miners to dedicate their computing power to training machine learning algorithms on that data. Currently, companies don’t have the budget for algorithm training infrastructure. However, they are willing to pay more than the fees miners currently own on cryptocurrencies. Neuromation spans the gap between those opportunities.
5. Why is your industry ripe for disruption?
MP: Every AI project has two major components: data (modern complex models require huge labeled datasets to train) and computational power (a single experiment may require weeks of running time on a high-end GPU, and any serious project requires hundreds of experiments); here at Neuromation we are disrupting the industry in both directions. On the data side, we are promoting the idea of synthetic data: e.g., to train computer vision models we create 3D renderings of scenes with the desired objects and place them into different virtual environments, which gives an unlimited supply of perfectly labeled images at a fixed upfront cost of manual labor. As for the computational power, we are developing a platform that will disrupt the industry by democratizing the training of modern deep neural networks. AI researchers and practitioners need thousands of hours of high-end GPUs, and we at Neuromation are harnessing the computational power of cryptocurrency miners (whose returns on investment decrease as more and more coins are mined), redirecting them through the Neuromation platform to training AI models, especially deep neural networks. The gap between what miners make from their hardware and rent prices in cloud-based services is 10-20x, and by bridging this gap we are disrupting the AI industry, addressing the two most important issues that any AI project faces: data and computational resources.
6. What’s the future of your industry?
MP: Over the next few years, we will see:
Prediction #1: A rise of specialized hardware for AI — proprietary Google TPUs and recent news from Bitmain are only the first examples, and soon we will see many chips designed specifically for training neural networks that will bring deep learning to new heights;
Prediction #2: A centralization and at the same time democratization of AI research: new specialized hardware will probably concentrate mostly in large-scale clouds (alas, the days when you can stay on the bleeding edge of AI research with an off-the-shelf GPU in your garage will pass) but due to solutions such as the Neuromation platform the rent will be cheaper and more accessible than ever.
Prediction #3: Commoditization of basic AI models and services: we already have some commodity datasets (e.g., ImageNet in computer vision) and models trained on them available for the general public, and in the near future this trend will bring a lot of “basic components” that an AI researcher will be able to work with and combine directly, without tedious fine-tuning; this will be partially a technical process of making what we have easily accessible but will also require some new theoretical insights.