Well Alexa needs to be updated for Pi4, it does not work yet.
Still based on Stretch?
I have no idea what hardware is in those smart home hubs.
https://smarthome.university/home-assis ... assistant/
But the Pi4 is probably now over the line in the capability department.
So do you buy a "Smart Home" setup that could be hacked or make your own with Pi's?
Anyone could have an Ironman JARVIS?
I have compiled ARM's Compute Library on 3's and 4's.
Still lots of other things like TensorFlow, Caffe, Leaf etc to try.
Some of these had issues with the older Pi's but just compile now.
Will full TensorFlow 2.0 work as well as Lite?
TensorFlow is tested and supported on the following 64-bit systems:
64 bit only?
Computer Vision now has more power, enough to do more complex stuff locally?
AL/MM/NN/CV etc software libraries all get a boost and are much closer to mainstream Linux.
Networked AI is where the models are trained somewhere else on big machines then run locally.
10,000 images is normal for training to teach a computer the difference between a cat and dog.
Once that model is learned then it could run on a local Pi.
Think of it as training the brains neurons to tell cats and dogs apart.
Once that "Neural network" has learned that "brain" can be copied to a local Pi.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges