AI inference engines. The learning part is done on big machines in the cloud. The inference part (giving answers) can be done on edge devices. There's a ton of articles listing the competitors, here's one fairly at random: https://www.aiacceleratorinstitute.com/top-20-chips-choice-2... (Note the article covers both learning and inference chips.) What's interesting is I can't find out much about Sony's chips.
On devices like the Nvidia Jetson Nano you can also train models.
I think Raspberry Pi will move in this direction.
The Jetson Nano has replaced the Raspberry Pis on things like the Donkeycar project. I guess they don't want to become obsolete.
Depends on the marketing department in question
If we're talking about Sony, it's probably vision related A.I. accelerators. Their Alpha cameras depend on them very much. Even my now old A7-III can do amazing things realtime.
Chips optimized to perform the type of calculations used for NN inference at high parallelism. A good example would be the google spinoff https://coral.ai/ (though their usecase is highly limited by sub-par software constraints)
Chips with some sort of neural network accelerator or related hardware.
no idea but if you asked me what I would want it to do is tensors for GPT and DL image processing inference and if possible training (so I can at least finetune a pretrained model)