What about asking Google Voice Actions to send a text message, read your messages, when the birthday of a contact is, set a timer for you, remind you of something when you leave home, play a track, etc. They compared only questions Google VA could handle and left out 90% of what makes Siri awesome.
You can send text messages, set alarms, and play music via Voice Actions, but it's not as polished as the Siri experience. I can't get it to read emails or SMS for me, and geofencing is nonexistent.
using 2.3.4, i can't _really_ play music. i can say "play the beatles" and it will offer to search youtube for the beatles, but it won't search the music on my phone.
You got to say "listen to [song/artist/album]". Which, of course, shows one place Siri's got the advantage currently, but I am sure Google is working on that now.
it still won't play my music. i get "complete action with [youtube|subsonic]", but no option to actually play the music on my phone, using the stock music player.
I think it worth pointing out that Siri's achievements beyond google should be credited to the startup that created it. It's popular and fun to give Apple all the credit, and they made a great purchase.
But this is not a story about Apple the innovator. This is a story about a startup that won big. That they were bought by Apple and integrated instead of remaining a strong new company is more a reflection on the current state of the software business than anything else.
All credit to the Siri team and the projects that it is derived from. The funding for this includes government (DARPA) as well as several academic institutes:
"Carnegie Mellon University, the University of Massachusetts, the University of Rochester, the Institute for Human and Machine Cognition, Oregon State University, the University of Southern California, and Stanford University, as well as from SRI."
Siri's personality does one thing Google Voice Actions doesn't: develop an emotional connection with the user. I fully expect Siri to impact the number of people switching to Android. It would be less like switching phones and more like firing you personal assistant.
I guess it depends on what type of person you are. I'd rather not pretend to be having a real conversation with an electronic device and would rather give simple voice commands when needed.
That being said the more conversational style, with the emotional connection as you put it, could easily be added into the Google Voice actions as well I'm sure.
It's been my experience that Siri responds equally well to curt commands as to colloquial phrases, which is how voice commands should be (and how voice commands to real assistants work, somewhat.)
The future will tell, but I would bet that all those human-like interfaces are gadgets deemed to stay gadgets. Would you voice drive a car? a plane? machines interface can use metaphors, like desktop and folders, but those are just metaphors, ie a way to mentalize a tool, which is much better controlled through direct orders than "conversation".
I 'voice drive' cars all the time. I get into the taxi and say "take me to the airport". The only reason I care about the steering wheel is that robots aren't very good at driving yet. Someday, if robotic technology improves to the point where it is as safe(or safer), that's the way we will do it 90% of the time.
Of course, there will always be cars with steering wheels, just like we still have horse carriages. You can still paint with a brush even though photoshop and illustrator exists. If there were a siri-controlled TV (and there probably will be) it will still come with a remote control.
Now that I am on a keyboard I can try to explain my thoughts a bit further: Siri has two features that attract attention, but they could also lead to failure.
First, the use of the sound channel to control the machine. At first, it seems a good idea, but sound is notoriously hard to parse for machines. Heck, even for humans: I have a friend who is deaf of one ear, and he can't follow a conversation in a crowded room, not because of remaining ear's accuracy, which is very good, but because we need 1) two ears 2) a head in between and 3) a serie of tiny head moves to localize accurately sounds in our mental 3D reconstruction of the surrounding world.
Sound channel seem convenient as it transmits information without wire to the recipients, but there are not so many examples of a good controlling of a machine by sound channel. (Humans and animals are NOT machines, sorry to insist.) Other wireless channels may prove much more efficient, infrared is a common one, and maybe brain waves will work one day. I don't say sound channel is useless: Sound channel is very good at delivering emergency feedback to the human controller, for example.
Second, the human so-called natural language (which is actually purely artificial). This is one of the greatest invention of humankind and it is extremely powerful for conveying information between human beings, when they share a common language. Moreover, it fits extremely well with the needs and powers of human brains. But again, it would by a sin of anthropomorphism to believe that machines should use mainly natural language to get their instructions from human beings. Machines have their own needs and powers. A machine can beat the best on a chess board, but is still unable to tell a cat from a dog on a picture, which my 2-years old little boy does easilly, and for fun. Machine are different, and making them human-like is deceptive. Therefore Man-machine intraction is not to be modelized on Man-man interaction (even if, as stated above, a human-like metaphor may help).
I have been hired to work on natural language during Bubble 1.0, and it was already widely known that this thing doesn't work, is not efficient, and is not even what users really need.
Now let's see it from another angle. I had this morning a task to achieve: fix a broken window lifter in my car, in Beijing. It is not so easy. I first did find Suzuki's Chinese translation (lingmu), then searched "lingmu fix car" (in Chinese characters) on my phone's mapping app, then I did choose one workshop not too far away from my way to office, then I clicked "call" and called the provider to make sure they had the broken part, then clicked "direction" to get navigation direction to the place. I worked perfectly, I felt thankful to all who have made this possible. But I don't see where a Siri-like assistant, or even simple sound input, would have changed the game. Voice input could have been helpful during the "fill the search field" step, but if you know a bit about Chinese, you'll know that when typing Chinese you need to first fill the sound, and then choose the characters in a list. Maybe a Siri-like assistant could have helped me clicking on the "get direction" button, but this was not the hard part. The step where I chose the car fixing shop among ten or twelve is more interesting. I don't think too much intelligence in the process would help: even if the assistant would be able to ponder all parameters (traffic, distance, etc.) I would still want to choose myself in a list. Here, the most effective could be a map projection on the windscreen and an eye-tracker with a "brain click".
Actually, I think "brain click" has a lot of unveiled possibilites, and controlling a phone with that would also avoid adding to the overall noise, which is a big contemporary problem, and which I did not help too much writing this comment.
Has anybody seen a comparison of actual speech-to-text capabilities, preferably done in a proper manner (using speech recordings of different accents, etc)?
