Who needs voice controlled email? Anybody who could use a human secretary. The 'human' part can be replaced with "relatively intelligent", couldn't it?
90% of the time I use my e-mail, I'm around other people and talking to my mail client would range from annoying and intruding on other people to breach of confidentiality.
The remaining 10% maybe it'd be useful, but it'd need to be useful enough for me to learn two separate ways of doing things - one with the spoken interface, and one without.
We are intentionally focusing in those 10% first, because email is a very broad use case. For now, you can think of it as a tool to catch up on email while driving or on the run.
I work in a startup that does zero knowledge email, so I think about stuff like this a lot. Would it be possible to do the work at "runtime" i.e. process the emails on the user's device?
I highly doubt it. I'd imagine the training data is several gigabytes if not terrabytes of corrospondance.
I don't know much about this type of AI work, but I can tell you a little about the scale of things where I do work.
I work with kana/kanji conversion and it is a similar problem as just getting the phonetics part of speech to text sorted out. First you convert the sound signal to phonetics, then you need to convert the phonetics to actual sentences. A surprising amount of English sentences have the same phonetics, so it is a non-trivial task to make sense of these. Try saying "Recognize Speech" and "Wreck a nice beach" out loud.
I'd dare say Japanese is even worse, so a lot of data is used to infer what the writer means with the phonetic input in kana/kanji conversion.
Baidu uses 2.5 terrabyte[1] of data for their kana/kanji conversion system, which has both an online and offline part, where the online part makes use of all the data and boosts performance considerably.
This is an interesting perspective. Do you use a home-rolled email client/server, and only exchange emails with other people with a similar setup? Completely avoiding "third party apps" with respect to email seems like it would take quite a bit of effort.
Interesting concept. I did sign up to test it out. However, I think that voice control is not practical with the amount of email we receive on daily basis. Some people will receive more than 100 messages per day that needs to be answered and a voice activated software will not cut it.
However, looking beyond this challenge, the idea is still cool and having an AI assistant to sort out your inbox will be certainly a breakthrough technology. Besides web, email is perhaps one of the greatest innovations of the 20th century but it needs to be brought in the 21st, i.e Email 2.0.
You are right, it doesn't have to be a voice-only interface. rather, voice UI is a starting point for us, since that is a use case we already understand well. But it could work just as well with written/texted instructions to your assistant.
After 5 seconds I thought the video was an ironic comment on our culture of constant interruption but it's not.
You have to put a lot of faith into that software if you actually use it while driving because checking if the answer is correctly chosen with the phone hanging over the wheel is a big no-no in my driver's book (it does beat actually typing out the message with one hand but that is already an even bigger no-no).
Semi-automatic voice activated replies is something I'd like to try out to see how it feels though. I am curious how the experience would work out considering I am more of pushing buttons guy.
I also don't get push notifications of my e-mail. I also don't get a ton, but it's a well known fact to my colleagues that I don't reply to e-mail instantly and if they're sending me an e-mail expecting an immediate reply, then they should re-evaluate why they're sending me an e-mail.
Do I have to keep the phone unlocked and the app open to have the kind of experience demonstrated in the video? If that's the case, given the kind of batteries in the smartphones, the fun won't last very long. I can keep it connected in my office, I believe.
That looks neat in a video. But as the person that receives the email, I would not feel valued at all by the thought that Joe doesn't even take the time to reply but instead lets a robot reply for him.
If you don't want to take the time to reply, then don't bother sending me anything at all. And it's not about having a crazy schedule. We have time. In fact, we have more time than ever before because robots already automate repetitive tasks. But a human interaction is not a repetitive task IMO.
This is something I think will be a cultural shift over time. There is something less authentic about it, but I think the "sender" still gets credit for the message. In some ways, they get more credit since the reply should be more timely.
I regularly schedule meetings through people's assistants and they probably reply to my emails to, but I don't know about it. This will be the same thing.
On a sleepless night, I was playing with my iPhone and accidentally activated Siri. I never use Siri but I was bored and the screen had a couple of suggestions, one of them being something like "Is my daughter home". Puzzled about what that could possibly make Siri do (I don't have a daughter), I asked her that. Excited about the sudden attention, Siri decided to call a person in my contacts, a guy that I barely know and only spoke with years ago. Panicking, I pressed the home button instead of hanging up immediately, so the call probably went through, possibly waking him up in the middle of the night. I never heard from him so I don't know. But the experience was rather unpleasant.
I think the demos in this video look neat but I don't see myself using AI for anything that I have to rely on any time soon. I would constantly feel the need to double check it anyway.
What's up with that specific kind of really upbeat music in promotional videos for IT gadgets and software? Everyone uses the same jingle composer/ukulele player?
The one thing that absolutely keeps me from using most of these services SIRI, Now etc...to their fullest is the voice commands. Not because they aren't up to the task, which they often aren't, but because I don't want to announce everything I am doing to the world.
This takes that to even another level of 'nope' for me by reading my emails aloud to whoever is around. The video example is exactly the stuff that I wouldn't want other people hearing.
this looks awesome, but how do you deal with noise? If the use case is being able to process email when not explicitly doing that task, wouldnt it essentially be when you're waiting for the train or playing with kids or even around a noisy office?
I have a tough enough time with IVR systems today because of my kids talking in the background that I have to switch to touch tone mode.
Valid point, but speech recognition on mobile is improving very rapidly - e.g., see Robin - our other Android app. Also, email messages give you the context which makes it easier for the NLU engine to extract the meaning of your utterances.
If the makers of this software are on HN, I am quadriplegic and this would make my life about 333.25% easier. For realz. Seriously though, if I could navigate my email with my voice that would totally negate any accessibility issues I have with email clients.
I can use email fairly successfully, Siri does a pretty good job but if this works as advertised I would definitely love to try it.
56 comments
[ 2.4 ms ] story [ 71.3 ms ] threadThe remaining 10% maybe it'd be useful, but it'd need to be useful enough for me to learn two separate ways of doing things - one with the spoken interface, and one without.
I don't know much about this type of AI work, but I can tell you a little about the scale of things where I do work.
I work with kana/kanji conversion and it is a similar problem as just getting the phonetics part of speech to text sorted out. First you convert the sound signal to phonetics, then you need to convert the phonetics to actual sentences. A surprising amount of English sentences have the same phonetics, so it is a non-trivial task to make sense of these. Try saying "Recognize Speech" and "Wreck a nice beach" out loud.
I'd dare say Japanese is even worse, so a lot of data is used to infer what the writer means with the phonetic input in kana/kanji conversion.
Baidu uses 2.5 terrabyte[1] of data for their kana/kanji conversion system, which has both an online and offline part, where the online part makes use of all the data and boosts performance considerably.
[1] http://www.aclweb.org/anthology/I/I13/I13-1172.pdf
However, looking beyond this challenge, the idea is still cool and having an AI assistant to sort out your inbox will be certainly a breakthrough technology. Besides web, email is perhaps one of the greatest innovations of the 20th century but it needs to be brought in the 21st, i.e Email 2.0.
You have to put a lot of faith into that software if you actually use it while driving because checking if the answer is correctly chosen with the phone hanging over the wheel is a big no-no in my driver's book (it does beat actually typing out the message with one hand but that is already an even bigger no-no).
Semi-automatic voice activated replies is something I'd like to try out to see how it feels though. I am curious how the experience would work out considering I am more of pushing buttons guy.
What I like about my email is that I don't have any push notifications from it. I check it when I want to.
Granted I'm not the kind of person who gets a lot of emails and thus would benefit from an AI email assistant.
Next up: "Tell Zuck I can't make his birthday party"
I thought it was a joke product too.
Does it work with handsfree?
Joe> Keep responding that I'm working on it
Bob> Keep reminding Joe that the report is urgent each time he responds
If you don't want to take the time to reply, then don't bother sending me anything at all. And it's not about having a crazy schedule. We have time. In fact, we have more time than ever before because robots already automate repetitive tasks. But a human interaction is not a repetitive task IMO.
I regularly schedule meetings through people's assistants and they probably reply to my emails to, but I don't know about it. This will be the same thing.
I think the demos in this video look neat but I don't see myself using AI for anything that I have to rely on any time soon. I would constantly feel the need to double check it anyway.
Companies like to pretend that they have some magic AI that makes things just work but we simply aren't there yet, and far from it.
This takes that to even another level of 'nope' for me by reading my emails aloud to whoever is around. The video example is exactly the stuff that I wouldn't want other people hearing.
I have a tough enough time with IVR systems today because of my kids talking in the background that I have to switch to touch tone mode.
How does this work in the real world?
I can use email fairly successfully, Siri does a pretty good job but if this works as advertised I would definitely love to try it.
Thanks.
(beware their site is based on frames)