3 ms·
Something in that spirit is already making progress: compute attached to cell phone towers, called "edge computing". Because you are pooling computer resources
by david-gpu 2y ago
Something in that spirit is already making progress: compute attached to cell phone towers, called "edge computing".
Because you are pooling computer resources across many more users you have better amortized cost than if you had a workstation on every home that was idle 99% of the time. Of course, latency and bandwidth to the cell tower is worse than wifi, but better than if the compute is done on a remote server.
I have no idea of which model will succeed for which use cases, but the idea is sound.
- giancarlostoro 2y agoNot sure how well this would scale if every users needing to compute all at once. Which all it takes is some meme new app that requires heavy computation and then everyone downloads it.
- david-gpu 2y agoIt has the same scalability issues and the same scalability solutions as cell towers themselves, doesn't it? Densely populated areas have more cell towers and thus smaller cells. Again: which computation paradigm works best depends on the use case. It is not an all or nothing situation.
- nthingtohide 2y agoSmart mirror, smart appliances, smart robots (of various kinds), all could benefit from such a central compute server which can work without internet. All appliances will have voice interface, and will require heavy intelligence so it is better if the model is downloaded to a central place at home. This model also means all home appliances can be equally smart with no restriction on form factor of the appliances themselves. Nvidia could model itself around current mobile companies selling OS, models, everything in a nice compute-rack package.