It's isn't just branding though, if I'm right. The wake word needs to be something that is easy to pick out from a complex audio environment and not be a regular part of common speech so cause confusion (computer fails a bit in that regard) so allowing completely custom wake words might cause reliability issues. Also some might choose wake words that the brands don't want to be seen listening for: I might choose "slave" as a Blake's 7 reference for instance but that could easily offend some if they overhear, and there are many other epithets, slurs, and swears, that would not be deemed suitable either.
Though it can get a little confused occasionally when I'm playing some variety of Star Trek on the nearby TV.
I think Thom would approve.
I don't see why the system could not be designed such that one could submit a wake-word, which is analyzed for suitability in terms of being sufficiently distinct in complex audio environments, and then checked against a blacklist.
Shouldn't be beyond Google's technical capabilities.
The thing is, a LOT of people avoid Google Home/Assistant because they simply can't bring themselves to blurt out cheesy branding like that.
They really ought to solve that problem. Probably they will right about the time they port Inbox bundling to Gmail and provide a Drive sync client for Linux (it's been, a decade?).
Unfortunately "enumerating the bad" is a strategy that never survives contact with the public.
They could also walk around saying offensive words when not addressing the computer.
Or is google planning on 'fixing' that too?
:)
Unfortunately some would reprogram other people's devices to cause offence for a jape and some will get offended and blame the company for letting it happen.
The difference from Google/Amazon/Apple/MS's point of view is that if I run around yelling the N word and their device doesn't even notice then they can't be seen to be complicit by unreasonable people. It would just be me being an arse.
Heck, some would get offended at the possibility of one of a company's devices responding to loaded words like that even if it never actually happened.
Of course there is another reason: having people use the same word re-enforces brand recognition even if the wake word is not the actual brand just a word/name people associate with it.
It's honestly the reason for me.
Such a minor thing, and it's super annoying. Google lost out on hundreds of dollars from me on this simple change.
And the resulting announcements. "INFORMATION: Four minutes have elapsed." for a timer, etc.
I’d buy a big subwoofer to make the INFORMATION response shake the house too!
Google's Mid-2000s marketing and product-nomenclature was on-point though:
* Google Mail
* Google Maps
* Google Image Search
I refuse to use a service that requires me to say its globally recognized name so often I will probably become brainwashed to it. And then there's the older hacks with TV commercials that took advantage of those defaults, and the (cooler) hypersonic transmitted voice command attacks, or the ones delivered by vibrating the device's microphone with a laser, etc.
None of these attacks would have worked if the product trigger wasn't so predictable from the get-go.
Eventually even Raspberry Pi stopped using the default pi/raspberry default combination. How we invoke our voice-activated programs should be treated with equal care.
Or maybe not.
/s /h
That idea was canned...
True story.
I think he didn't want to dilute the brand, and have the association front of mind or something, but the people on the alexa project managed to convince him to go with alexa instead (as it's confusing, and arguably a better name).
I think that's part of the reason for the different wake words, one of which is amazon (though the confusion with people called alexa is likely a much bigger one. It's probably why they have amazon in that list at least).
I can hear it now “We finally have a use for that brand we spent all that money on..”
https://www.techspot.com/news/46668-googles-answer-to-siri-m...
Say "McDonalds!" to end the commercial.
https://www.fastcompany.com/90185994/sony-files-patent-to-ma...
Because sure sounds like it is.
It sounded dumb enough for two adults to say it all day, but the convenience was worth it. Having my little one start using the phrase to talk to things is definitely where I draw the line.
Besides, now that it tries to make suggestions after every other command all you really hear in my house in regards to them is "hey google shut the fuck up and do what you're told". Unfortunately the stupid puck doesn't allow you to interrupt the unsolicited FAQ any longer.
Come to think of it, I think I'm getting rid of most of these things this weekend.
Of course Amazon isn't likely to ever do such a thing, because god forbid we don't have the ability to suddenly order something from Amazon on every device at all times.
Comically, that's actually one of the things its the worst at. I tried to do it exactly once and swore I would never do it again. In the time it takes to read the first search result, I could have found and ordered what I want from my phone. If that's what they were going for, I can't believe they didn't scrap the project.
Supposedly they're fairly good at re-ordering, but given how often Alexa mis-hears me, I'm not a fan of doing anything involving money on it.
Also, I've mapped all of the functions we use to "recipes" that include the word "please" so that they don't work unless you use the word "please." i.e. the upstairs lights are named as for example "the Chicago lights" and there is a recipe that says "turn on the Chicago lights" when someone utters "Computer, please turn on the upstairs lights." So if you say "Computer, turn on the upstairs lights" it says "I don't know what upstairs lights are."
The disadvantage of this approach is that you need to get the string exactly as stored.
"#Karma.. Police.. Arrest this man, he talks in maths...#"
There are, unfortunately, millions of ways we form relationships with brands. A brand's reason for existing is to live somewhere in our psyche either subconsciously or consciously. We say "Q-tip" instead of cotton swab, "band-aid" instead of bandage, "Advil" when we need ibuprofen, or "Tylenol" when we want acetaminophen. When we see white polar bears around Christmas time, we think of Coca-Cola, and so on and so forth.
All that to say, kudos for you for taking a principled approach, but I am not sure it's going to achieve much.
I've been meaning to have a play with https://mycroft.ai/ but only out of interest, I don't see myself leaving it running.
Now you can mispronounce it, that’s the only leeway we get.
Not a bad option. I would much prefer asking 19th century Russian author, Nikolai Gogol, to turn on the lights.
"I am not a number I am a free man!"
Last year my parents rented a room at a 5 star hotel. The hotel had a smart mirror and I only ever saw that mirror play ads. It's insane.
Man. Seems kinda relentless.
Submit a PR that responds to "Ok Google, enough!" by degoogling your life including automatically submitting your resignation.
"Oh, let me demonstrate this. Ok Google, turn off the lights."
"Ok Google, turn on the lights now."
"Ok Google, mute."
"Ok Google, turn on the lights."
"Ok Google, turn on the lights, damn it."
"Ok Google, turn, on, the, lights---there you go. I swear it works better yesterday."
Might as well just flip the switch myself, if I have to debate the assistant half of the time in total darkness.
"Alexa, ask phillips hue to turn on the living room light"
vs
clap clap
In Star Trek they were perfectly comfortable saying "Computer! Do the thing" in a more specific, 'computer' intonation. It was all fairly natural language, but there is no attempt to pretend the computer is a person. This made the thing feel more futuristic than what they're trying to do now.
I am flabbergasted that the following hasn’t been an option:
- hey Siri
- yes?
- what are the last three releases from <artist>?
- X Y and Z
- search again without EPs
- W X and Z
- Play the first one
- <playing W>
- thank you Siri
<conversation closed>
Also, with attention tracking that -already- exists with the FaceID array, the phone can know when it is addressed and when it’s not. You know, just like when you’re talking to someone, you usually look at them...
I have a convenient, quiet, perfectly reliable single-bit computer on the wall in every room in my house. It costs $1, can be operated by a 2 year old, is perfectly intuitive, and the only downside I can think of is that occasionally it necessitates I be very slightly less lazy than I might otherwise be able to aspire to, i.e. when I need to go downstairs and turn off a forgotten light.
Considering my house is all using LED lights on hydro power, it's probably better for me to just leave a 5w light on all night than it is to install a Google Home setup here, in terms of my carbon footprint.
That is the key point, for me anyway. "Computer, all lights off" instead of getting out of bed when I've left something on is one of my lazy wins from voice control. Setting multiple timers without touching things with messy hands while faffing in the kitchen is another.
-- is Google misinterpreting your voice? E.g. does it hear a sound it thinks is "play" in the middle of your phrase?
-- or is it some weird statistical model that because of invisible and irrelevant correlations, sometimes concludes it's more likely you're asking for music? Like the song with that title is currently in the top 40, or was played by you in the past, or something?
(My brother in law is Australian living in the US, and has to use his "sarcastically fake American accent" to be understood on the phone. He, more politely than I, calls it his "phone voice". It's the same voice all my friends here use when parodying American stereotypes... I bet it works on Ok Google too.)
From a quick search, it looks like national accents are less of a problem, and the real difficulties like in dialects and foreign accents (e.g. a Spanish person speaking English).
But the idea that the training corpus used for speech recognition models is all "20-30 year old white male techbro bay area" guys is ludicrous. Where are you even getting this? Speech corpuses used for testing are put together by people who take demographic diversity seriously. On the other hand, they are sometimes biased towards a single mainstream national dialect. (E.g. "standard" American English, as opposed to a thick Boston accent.)
[1] https://voicebot.ai/2018/09/13/google-assistant-takes-lead-i...
I'm not entirely convinced Google test _anything_.
Certainly not convinced enough to not play it up in the context of a snarky dis on Google and Googler culture.
Don't you recall the infamous fifty-shades-of-blue thing with Google? Google's whole thing is that they quantitatively test everything.
Maybe you've been living under a rock the past 20 years? :P
Googler #1: "Marissa want to change the link colors _again!_ What can we do to make her stop micromanaging shit like this?"
Googler #2: "All we need to do is convince her we are obviously wrong about 'graphic design-y things' because we're developers, and that she can prove we are wrong by making a 'data driven' decision. Then we'll say 'The only way to prove we are wrong would be an insane global 0 way multivariate test of all possible colours!!!' and she'll _insist_ on it. Then we build a tool and run the test and point at the results every time she tries to make us change it again. And we can offer to run the tool for _everything else_ she micromanages us about in future too!"
Googler #1: "Brilliant! So, do we build the tool and run the test? Or just make up the results and present them to her?"
Can confirm, in high school I had to occasionally go AFK after being beckoned talk to the amazon echo because it wouldn't understand my parents' accents.
A lot of immigrant children get used to translating for their parents, but repeating commands but with a native accent is a newer phenomenon.
Used to be the same in the store, tuna sandwich pls. Huh? What? Oh, toona!
Settings > Siri > Language > English (United Kingdom)
Settings > Siri > Siri Voice > British
Apple advertises:
> Since Siri is designed to recognize accents and dialects of the supported countries or regions, Siri’s accuracy rate will be highest for native speakers.
Or is this problem just with American friends' devices?
My Siri's set to American but when I travel to London local maps and stuff all works fine there?
Hey Pal! for Edinburgh
Hey Jimmy! for Glasgow
Yo! for LA
Oi! for NY
Doode for SF
etc
80% of the time my lights would turn on.
20% of the time, I’d be greeted with: “Ok, playing ‘Turn on the Lights’ by Future on Spotify.”
And I’d stand there in the dark, listening to music I don’t like, questioning my life decisions.