On: Introducing Transfer Appliance: Sneakernet for the...

[twakefield]

"Like many organizations we talk to, you probably have large amounts of data that you want to use to train machine learning models."

I understand's Google's bias here but doesn't it usually make more sense to bring the programs/models to where the data already is?

[btian → twakefield]

GCP has proprietary machine learning accelerator hardware that you can't buy.

[bpicolo → twakefield]

The problem isn't the programs, it's the hardware. GCP has specialized / scalable hardware to build models on.

[zeroxfe → twakefield]

It is extremely expensive to setup and maintain infrastructure to process big data, especially when you get into the petabytes.