AI companies to reuse crypto mining farms for deep learning in healthcare
Longenesis, Ltd, a Hong Kong-based leader in applying the latest advances in deep learning and blockchain technologies to healthcare announced a research collaboration agreement with Neuromation, Ltd, a Tallinn, Estonia-based company focused on transforming the cryptographic mining infrastructure into a massive cloud for deep learning applications. Neuromation recently successfully raised over $70 million through a token generating event (TGE) to develop a system for generation of the synthetic data using the deep neural networks trained on a large number of examples and using the processing power of the many graphics processing units (GPUs) in the blockchain community.
"Here at Neuromation, we are developing a platform that will democratize AI development by making the computational resources more accessible to AI practitioners; the computational power will come mostly from cryptocurrency miners wishing to earn more from their GPU farms and/or diversify their business. But even if you do have the computational power, you still have to have the data to train machine learning models on", said Maxim Prasolov, the CEO of Neuromation.
Both companies are developing global ecosystems using blockchain technologies. The approach utilized by Longenesis was recently published in a peer-reviewed journal. The is developing a secure ecosystem to allow the individuals to take control over their life data and profit from selling or licensing it to the application developers. The application developers need access to the high-performance computing solutions to be able to train the deep neural networks that can be used to provide the valuable health-related services back to the individuals or to discover new life-saving therapy. Neuromation aspires to become one of the largest providers of the computing infrastructure for deep learning by repurposing the enormous installed base of the existing mining infrastructure.
"We are excited to partner with Neuromation. We previously published with their CSO and hold their research capabilities in high regard. Our collaboration is two-fold. First, to train the deep neural networks we need processing power and we are looking to repurpose the cryptocurrency mining equipment to do that. Second, to generate synthetic data, one must train on large data sets of real-world data and have balanced and diverse data sets, so Neuromation is an obvious consumer of data and can be a part of the marketplace", said Alex Zhavoronkov, PhD, CSO of Longenesis Limited.
"The primary aim of Longenesis is to create a global data marketplace that will help provide personalized but at the same time secure data for healthcare application developers. Hence, the missions of Neuromation and Longenesis are perfectly complementary to each other. Together we will bring to AI practitioners in medicine and healthcare both the data and the computational power needed to train AI models and improve healthcare throughout the world. By announcing this partnership, we are laying the foundation for the future world of individualized healthcare based on artificial intelligence", said Sergey Nikolenko, PhD, the Chief Research Officer of Neuromation.
The companies agreed to bridge each the ecosystems to enable rapid expansion and allow the large corporations to procure a large amount of data from the individuals consenting to the specific uses of their data and to procure the high-performance computing infrastructure required to train the deep neural networks.