Keep away from U.S. AI and IoT frameworks!
Every larger U.S. company is trying to set standards in AI and IoT to finally make big money with _’training courses’_ (software itself mostly is free!), desperately to create a _’vendor lock-in’_ with specially adapted NN hardware, like Google with its TPU, Amazon with XilinX FPGA boards, Microsoft/Intel alliance with _’Movidius Neural Compute Sticks’_ , IBM with _Cloud Foundry_ , Dell/EMC with its new, ‘open’ _EdgeX_ IoT framework.
All larger U.S. companie now are heavily working on creating their own SDK, with Microsoft taking a leading role with VS and VSC to make AI, IoT frameworks as much as ‘consumable’ as possible, _’embracing, extending, finally extincting’_ all others.
In fact, sending encrypted data from a Micropython enabled Embedded Device (e.g. ESP32) into the cloud is just _10 lines of Python code client side_ and another _20 lines of Python code server side_ , see YouTube examples.
Creating your own (self learning CPU based, no special neural hardware neeeded!) _artificial intelligence_ with NumPy (or new, highly optimized ARM CMSIS and Compute libraries) and Python is just 11 lines of code, see: http://iamtrask.github.io/2015/07/12/basic-python-network/
It’s so dead simple, not even worth to read the manual of some AI / IoT frameworks.
And we all know since Oracle-Google $8.8 billion lawsuit, that APIs in U.S.A. now are ‘patentable’, kind of.
So, whatever ‘U.S. made’ framework you use, you’re legally trapped.
My urgent advice: Keep away from U.S. software! It’s all _artificially bloated’_ to give us the impression, that we can’t live without.