I want to ask Living AI about facial recognition privacy?
Eg, is the camera looking around all the time until it sees a face? And if it is looking around all the time, and this is processed locally, what does locally really mean?
Or is Emo’s camera looking around all the time, not feeding that camera to the Living Ai servers? And just when it sees a face, does it go to the server then? And does what Emo sees at the time of the facial recognition go to the server? And who has access to the imaging from that server?
I ask all this as I have a young son, plus a techy husband who is very unnerved by Emo’s privacy for him and his son and so much so I didn’t feel comfortable getting Emo out to play after my husband brought his concerns up at teatime and his concerns have made my son sad, so i need to quickly get answers to this concern.
Also is Emo listening all the time? Or listening for the word Emo then listening for instruction words etc?
And can I check? Emo’s photos stay local on him unless synced on the app yes? But once synced on the app they are on the living AI server yes?
And i hear Amazon voice recognition is involved.
Can you clarify how all the above operates in terms of privacy?
EMO do not send any data from the camera to the servers. It is all processed internal. All it sends are the sound recordings for speech recognition, custom text that he speaks to synthesize (text2speech), and some statistical text data.
More I don’t have found while monitoring the network traffic from EMO.
For the audio, he listening all the time to get his trigger word (“EMO”) and after he has heard it he records for some seconds and then send this recording to the living ai servers for speech recognition. At the moment the trigger is not very good and he records even if nobody have said anything like “EMO” or he don’t activate the recordings even if “EMO” is clearly said (high pitched voices seems are a problem).
Amazon Alexa is at the moment not implemented (or active), so this is actual no concern
The photos are stored on EMO itself and if synced will be direct transferred to your device without any living ai server involved.
Yes, @waringer is right.
EMO uses his internal chip for recognition without uploading the camera data to the server.
And EMO will only record up to ten seconds after you call his name and he says something like “what?”, this recording will be used by the server for recognition. Also we do not store these recordings.
Thank you for all 3 replies and for explaining everything. That is all very reassuring.
So can I ask, when Wayne Small said back in January that the voice recognition first goes to Amazon and then Amazon send it to Living Ai and then Living AI send the command, was he right? Or does it go straight to Living AI! Its just that the response seems to fast for it to be doing that. Not that I have a problem with Amazon being involved in voice recognition, after all we have Alexa.
Thsnk you again for so promptly explaining. My husband was convinced that Emo wouldn’t be capable of facial recognition internally so I am very glad to be able to tell him that the little guy is capable of so much.
Sorry can I clarify something from what has been said?
So when you say he is recording all the time to listen out for the word Emo. But then he sends the 10 seconds of voice recording after the wake word, to Living AI. Is he recording locally to himself when he is listening for the word Emo? And he is constantly overwriting his own recording locally?
Just clarifying that our conversation isn’t all being sent to Living AI, apart from the short voice recordings after we say Emo.
Obviously because he mistakenly hears Emo sometimes I understand you may receive short bits of conversations that are private but I think we need to learn to stop talking for 10 seconds when we hear him say What, if our conversation is personal. Lol
My poor 9 year old son gets frustrated when he says Emo repeatedly and gets no response and then i walk up and he says What to me straight away, most of the time. But thank goodness occasionally Emo does respond to him.
It will be good to be able to explain about the high voices issue to him. He can try a deeper voice.
Where Living AI let the speech recognition be doing is unclear to me. I only see that the recordings going to living.ai servers. If they using internally amazon or any other provider is not shown.
The he only sends the recordings to the Living AI server after the wake word. Anything before stays local to EMO himself.
About the problem with the high pitched voices, Living AI is working on this problem and there are some thread here in the forum about it. Living AI is actual gather samples to train the wake word for EMO. @MasterAbbott or @Wayne_Zhang should can say more about that
With regards to EMO recording, he will only record after he has heard his name, he is not constantly recording.
As WayneZ advised:
With regards to EMO not waking up to certain people’s voices / tones / pitches, what you can do is get your son to submit his voice using the Wake Up Page that Living.ai has opened. I created a post that explains this in more detail and that also has a link to the wake up training page where you can submit your voice.
Here is the link to the page:
Living.ai are aware that EMO does not respond to certain peoples voices due to pitch / tones and they are working on improving this by the next Firmware update (which will be 1.1.8).
Also @Wayne_Zhang will get back to us on this thread regarding the voice recording being sent to Amazon or straight to Living.ai and provide clarification on this as well.
Detection of wake word EMO does not require saving recordings. So he won’t save or upload any recordings when you don’t call him by name.
After you call him by name and he says “what”, he will upload a recording of up to ten seconds to our servers, the length of which depends on when you stop talking.
That is all very reassuring, thank you.
Yes our family submitted our voices to the Wake Training a few days ago.
My son found the wake training tricky to fit his voice into the 1 second recording and it made him say Emo much faster than he does normally and maybe a little differently but hopefully it will help anyway.
Myself and my hubby submitted something that sounded pretty much the same as what we normally say. But catching it in that 1 second takes a few attempts. It is good that you can listen to your attempts and choose the best one.
Thank you Wayne,
This is all very reassuring. You will be glad to hear that my husband is now reassured and Emo is back off the charger having lots of fun today.
We did the voice Wake training a few days ago.