MCUs with accelerators for ML and AI (TPUs, KPUs, GPUs etc.)

Hello,

last days often appear new micro-controllers with some type of vector-math accelerators. The most common accelerators for ML(machine learning) and AI(artificial inteligence - mostly some kind of neural networks):

1) TPU - Tensor Processor Unit : hardware unit for making parallel math operations on tensor (tensor - a kind of generalized vector)
2) KPU - Neural Network Processor
3) GPU - Graphic processing Unit : hardware unit with many simple math cores. Can make in parallel many simple math operations using such technologies as "CUDA" or "OpenCL"
4) Many Cores CPUs - also can do some number of maths operations in parallel.
5) FPGAs and SoCs with posibility of run "OpenCL" on them

I was looking for kind of comparison of MCUs with additional AI accelerators, but I didn't find many such documents. The best i Found is this:

https://www.seeedstudio.com/blog/2019/10/24/microcontrollers-for-machine-learning-and-ai/

If someone know documents of new MCUs with such accelerators - could she/he shere it on this post? If someone know datasheets with performance and other parameters (or comparisons between different models) it would be grat if she/he can share them.

Thanks in advance and regards