RE:RE:RE:RE:Aiking & oltimer , you think I’m reaching or
Their use-case is geared towards the cloud vendors and data center operators as no one really deploys large scale AI on-premise.
The conventional wisdom today is: ingest data from a variety of sources, on-premise, edge of the network, smart grids, factory floors, etc etc and then push the data to a centralized location(typically cloud) where you train/run the AI models.
To that end, their solution provides the means to reduce the cost of building/running this type of infrastructure:
- Removes bottlenecks for Linear Scalability
- Streamlined computational processes
- Lowers energy consumption
- Reduces the data center footprint
Note though, they're just a startup, Nvidia and several other large players own that space.
Regarding your question about whether DM would benefit from using their hardware...I would say not really. DM solution AFAK is cloud hosted for the most part, the gov agencies possibly run their NexaSmart instance on-premise out of security concern...although they could use a private cloud deployment. In any case, building such infrastructure is an unnecessary and costly capital expenditure and one most companies try to avoid...hence the shift towards OpEx model.
Perspectum tech stack (computer vision and image processing/recognition)...if it happened would nicely compliment DM's, especially with the cyber security business. The core of this technology is typically deep learning and convolutional neural networks (CNN)...which can be used in other domains. The same technology is used for self driving for instance.
I hope this adequately answers your question...