It's always frustrating reading these threads seeing how anecdotal experiences are given equivalence to systematic research. I also pity the children (or supervised employees) of the many posters who are count it a fail until an assistant reaches 100% accuracy.
Personally I've been using Siri pretty systematically since a couple years after it launched. I then added Alexa Echos in several rooms in the house due to the inexpensive cost and home-control flexibility.
Apple's stance on privacy and security is a big win for me, and while I find Alexa useful I'm terrified on how Amazon may be exploiting my data, or at least retaining things they shouldn't and potentially sharing with 3rd parties accidentally or intentionally.
All that said, I use Alexa as a first stop for home control since it's always on and available. It was much harder to set up than HomeKit, though they've improved that UI dramatically. It's much clumsier working with Hue lights in particular. But for Siri I must either activate my watch or have left a phone out within a 'hey siri' distance which reduces its utility.
Points in Alexa's favor:
- reasonable accuracy (vaguely estimate a 90% success rate, with fails correlating with lower voice levels, mumbling)
- works with more devices (particularly useful: Logitech Harmony universal remote)
- multiple wake word options (nice for segregating between different rooms)
- much cheaper options than HomePod for in-home access
- has a minimal response option
That last point is a VAST improvement over Siri, which insists on lengthy, chatty responses every time you control a device. That's cute the first 50 times or so, but gets TIRING after that. Echo simply gives a quiet beep in response. After all, if you just had it turn a light on, you can pretty much see if it worked even without hearing a response.
Points in Siri's favor:
- Apple watch 4 responses just as fast as using Alexa
- more sophisticated home control options (technically HomeKit, not Siri)
- better library of math options, including scientific units and physical constants (useful to me as an astronomer)
- higher accuracy (vaguely estimate >95% success rate in the tasks for which I also use Alexa)
- more voice choices
However the most important thing I've learned in the last few years is that digital assistants are a 2-way street. They have to get better in recognizing what you say, but you have to spend effort learning how to talk to them most effectively as well. Natural language processing is one of the HARDEST challenges in modern computing and personally I'm amazed by how GOOD things are already.
Not that there isn't always room for improvement.