Neuromation: Accurate Datasets for Improved AI
Neuromation have certainly created a bit of a buzz. Their token sale lasted only 8 hours, and in that time raised $71.7 million. So, what’s all the fuss about?
Data Set Generation
Neuromation are interested in the accuracy and level of sophistication of your data. By leveraging the computing power of miners and blockchain tech, they intend to create a large amount of datasets and deep learning models. A user can then either by a specific dataset, or task the platform to create a new one based on different parameters. Transactions will take place using their ERC-20 NTK tokens, and miners will be performing data generation and model training tasks as opposed to straight crypto mining.
At present, their usage examples are within the retail, industrial and biotech sectors. Using synthetic data, they have created accurately labeled datasets through which store managers can then anaylze live stock, including current shelf stock. This is the first live stage of Neuromation. They are also able to create datasets based on habit, meaning that stock requirements are predicted from data on customer habits.
Within the industrial sector, it’s all about real-time data. Again, using this type of dataset and labeling infrastructure, data is more quickly analyzed, and is up-to-date. This means that models such as drones can be trained in real-time to act based on immediate information. They will be rolling out a sandbox environment for model training.
Finally, for uses in the biotech field, Neuromation are proposing a lab that uses deep learning to lead to drug discoveries and improved biological systems. They are actively searching for partners in this area.
While these are the three areas that Neuromation are targeting at the present time, they have acknowledged that the clear potential of such datasets goes way beyond these fields. As a relatively new startup, they are looking for ideas and partners in order to expand the potential of data usage and AI training. And with a roadmap that shows the completed version of their platform as due for release in the middle of 2018, there will no doubt be significant interest over the next few months, especially considering the success of their high-profile ICO.