My AI algorithms can run on cheap consumer devices, which means that they could in theory be used on a large number of consumer devices to facilitate distributed, massively parallel machine learning and deep learning.
Obviously, this would expose researchers' data to the computers of ordinary people, which poses a security risk.
My question is whether researchers would accept this security risk in exchange for a lower price, together with a guarantee regarding the quality of execution (versus the quality of security).
Refer this link getting some ideas
https://www.sumologic.com/blog/machine-learning-deep-learning/
Some of the questions that I would have for any vendor in terms of such offering are:
1) How much is the lower price difference?
2) Can the users see where the data is distributed(for example by regions like amazon's) and can one limit the regions?
3) What additional performance increase besides distributed computation?
4) How does the cheap devices compare to Cloud provider's offerings in terms of performance?
5) Why not provide encrypted containers?
6) How do you track the distribution of machines to guarantee that my price is correct?
7) Do the computing devices that I use have concurrent "tenants" and if so how do the system provide fairness in process execution? How is the quality of execution guaranteed?
It all comes down to a cost benefit analysis with data that is not confidential nor proprietary (which limits the universe to a substantial potential consumer base, especially those looking for commercialization).
- Badges
- Science topic