A complete and unbiased comparison of the three most common Cloud Technologies for Machine Learning as a Service.
This is a surprisingly in-depth review of the Amazon, Google, and Microsoft’s offerings in cloud ML. It goes through core ML tasks like classification and regression and continues into specific services in speech, text, image, and video processing.
I don’t have direct experience deploying solutions like Amazon’s Polly in production applications, and I’m somewhat conflicted about this strategy. I’m very comfortable outsourcing compute and storage to a cloud provider, but outsourcing core algorithms to a black-box service worries me as a strategy for anything that’s going to be core to a product. There’s nothing fundamentally preventing you from switching from S3 to GCS, but can you really switch speech-to-text providers in the same way? If not, you subject yourself to serious vendor lock-in in a core part of your application.
I’m curious to hear any thoughts you have on this topic.