I wonder if this is a strategic steps following Amazon Echo's nice overtake on speech-based assistant area. Apple is really the pioneer on bringing speech based UI to the masses. Maybe they sense a threat?
With speech-to-text yet struggling on my iPhone to distinguish between what I said and what it figured I said.
Let's spare the ideology, I think the two of them are very much not yet there.
I'm elated though to learn that a meaningful university research got a honorable mention. There should be more searchlight on what's going on behind the closed doors of those hardworking varsity labs.
With speech-to-text yet struggling on my iPhone to distinguish between what I said and what it figured I said.
I have very very good success with speech-to-text. This is for everyday communication using iMessage, not for anything technical. The secret is that I have slightly modified my speech:
Just. Pretend. You. Are. William. Shatner. And. You. Simply. Speak. Staccato. (Ie, with very short pauses between words). Of course you also need good internet connectivity, because all the hard stuff is being done in the cloud.
I find that speaking into my phone is much faster than typing on the tiny keyboard.
Also, there are two parts involved with Siri:
- first is speech to text, parsing individual words
Applying AI to understand troves of persistent/temporary data archive threads in order to learn and reveal patterns and associations in complex datasets of patient encounters can help doctors prevent expensive unnecessary lab tests that need not happen again for another patient. That's conveniently doable for human-computer interaction.
The viscosity of natural language seems to be the insurmountable challenge here. Based on what I've experienced, visual perception by computers and translation between natural languages can permanently earn its success in a matter of time's length less than a decade. Not just speech-to-text because we still have to trick Siri for example by speaking slowly.
It's probably my angle, but it is not a qualifying success when we all yet have a lot of work underway to do.
9 comments
[ 0.14 ms ] story [ 40.8 ms ] threadLet's spare the ideology, I think the two of them are very much not yet there.
I'm elated though to learn that a meaningful university research got a honorable mention. There should be more searchlight on what's going on behind the closed doors of those hardworking varsity labs.
I have very very good success with speech-to-text. This is for everyday communication using iMessage, not for anything technical. The secret is that I have slightly modified my speech:
Just. Pretend. You. Are. William. Shatner. And. You. Simply. Speak. Staccato. (Ie, with very short pauses between words). Of course you also need good internet connectivity, because all the hard stuff is being done in the cloud.
I find that speaking into my phone is much faster than typing on the tiny keyboard.
Also, there are two parts involved with Siri:
- first is speech to text, parsing individual words
- second is the AI, what do those words mean?
The second part can sometimes be problematic.
That's exactly where everything goes dark.
Applying AI to understand troves of persistent/temporary data archive threads in order to learn and reveal patterns and associations in complex datasets of patient encounters can help doctors prevent expensive unnecessary lab tests that need not happen again for another patient. That's conveniently doable for human-computer interaction.
The viscosity of natural language seems to be the insurmountable challenge here. Based on what I've experienced, visual perception by computers and translation between natural languages can permanently earn its success in a matter of time's length less than a decade. Not just speech-to-text because we still have to trick Siri for example by speaking slowly.
It's probably my angle, but it is not a qualifying success when we all yet have a lot of work underway to do.
[1] https://en.wikipedia.org/wiki/Siri [2] https://en.wikipedia.org/wiki/Google_Now
https://www.google.com/patents/US7177798