Hacker News Reader

Comment by Barrin92 | Hacker News Reader

I think the first and last point are fairly straight forward, but I'm still a typing believer. I don't think that natural language or other interaction will replace the written word.

Typing is faster than speaking and less disruptive. I can't really imagine how everybody on a train in Delhi is going to talk to their digital assistants without a noise proof helmet on their head. (or going crazy)

Also we've found many ways to condense information in written form. I can write "IIRC" but it's hugely uncomfortable to speak this way. You can't use smilies or emojis either when talking. In a way natural language is much less feature rich.

View on HN

Once technology starts to touch the entire globe, and we start nudging farther and farther out in the long tail, we can't keep assuming that a single type of interface will be a good fit for each and every person. The interface is going the need to be split up into multiple pieces to accommodate many people from very different educational backgrounds and abilities.

It's like that Obama quote about the difference between government and tech companies "Government will never run the way Silicon Valley runs because, by definition, democracy is messy. This is a big, diverse country with a lot of interests and a lot of disparate points of view. And part of government’s job, by the way, is dealing with problems that nobody else wants to deal with."

> we can't keep assuming that a single type of interface will be a good fit for each and every person

Try arguing this with the NFB...

National Federation for the Blind?

National Film Bureau?

National Federation of Builders ?

National Facility for Biopharmaceuticals?

No Fucking Bueno

The first.

It's also more private. Not so much in the sense of secrecy, but I feel very self conscious talking to a machine in public. I feel like this aspect of ergonomics is often ignored.

I mean, this is a blogpost by Google, and their big push right now is voice assistants ¯\_(ツ)_/¯

I keep debating if this is ignored, or just that we're not the target demo.

It appears ridiculous that people involved in creating this voice controlled UX are not aware of the inherent limitations involved, but it seems like there is a massive gap between it's inherently beneficial uses that have emerged and it's an area we're all trying to get right.

Personally, I never wish to have a fully voice controlled experience. I could just be old, but voice control is literally almost never an option.

It is extremely difficult to type in Indian local language. I can hardly go to 10-20 wpm in Hindi or Bengali. Also as mentioned in the article more than 80% population in India is not at all comfortable in English. I have seen my mother (who has a master degree in her native language) struggling really hard to form a proper google query in English.

I don't know what is the correct solution but voice can definately be valid option

It's not terrible on Android - I use the Hindi keyboard to write Marathi just fine. It's slower than English for sure because Swype doesn't work (or at least I haven't tried and can't imagine it working the same way for Marathi). And also because I don't have as much practice with it as I do with English.

Mac OS also has transliteration tools built-in for typing Hindi and most other Indian languages (though, again, not Marathi - you have to use Hindi transliteration and hope you get the right word suggestions). You can use the Caps Lock key to quickly switch between English and Hindi. I don't know if Windows has a similar feature but I wouldn't be surprised if it did - it's even more popular in India than OS X.

google search works fine in hindi

https://www.google.com/search?q=गोलगप्पे

also typing in hindi using transliteration is as fast as english for me.

https://www.google.com/inputtools/try/

transliteration involves a good understanding of english alphabets and qwetry. Many people in india don't have that. Only people who regularly type using transliteration are those who are familiar in english

Also hindi language search is lacking many features which is taken for granted. Situation for other language are even worse

Look at the result for two same search in english and bengali https://www.google.co.in/search?q=delhi+to+calcutta https://www.google.co.in/search?q=%E0%A6%A6%E0%A6%BF%E0%A6%B...

> transliteration involves a good understanding of english alphabets and query. Many people in india don't have that.

That's true. I hadn't thought of it. OTOH transliteration tools are more for desktop/laptop users who use the keyboard that comes with their machine. On mobile (which is where most of these "next billion" users are going to be) it'll be a language-specific keyboard with no knowledge of English letters needed.

People claim that same about command line interfaces (and they are mostly right about them being more functional) but they are not the interface the 99% is going to operate their computer with.

Voice will be the next GUI.

>Voice will be the next GUI.

If we're talking really sophisticated interfaces like in "Her"[0] I agree. If the interface is working so flawlessly, it's difficult to see why you'd still want to occupy your hands with querying some knowledge base, composing blog posts/letters, etc.

However, I don't think we'll get there with current technology. When I think of the often awkward and cumbersome interactions I have with Siri, I find it really hard to imagine how this will evolve to a 'I don't need to worry about this at all anymore'-level in the next 5, 10, maybe even 20 years.

I suspect the giants of today won't be around anymore when truly voice-controlled interfaces come around.

[0] http://www.imdb.com/title/tt1798709/?ref_=fn_al_tt_1

...it's difficult to see why you'd still want to occupy your hands...

Bandwidth. You have several degrees of freedom with each hand, with each finger. You have one linear stream with voice.

Latency. You can flick a switch in a few milliseconds, but saying "turn off the lights" or "lights off" takes half a second or more.

Privacy. You can overhear a voice command. You can't overhear (very easily) a buttonpress or touchscreen swipe.

Accuracy. Even with perfect voice transcription, people misspeak easily more than they mistype. And mistyping can be corrected within a few letters, while misspeaking will require interrupting the stream to switch the voice UI into editing mode or something.

See, but you're still thinking in current voice-interface terms. If I want to edit something that I "misspelled" (if this will even be a problem, since you are _talking_) I'll just tell my digital secretary "errr, I meant <x>" and it will know what to do.

Same for turning off the lights: Sure, it's faster if you only consider the flicking of the switch. It's another thing if you also incorporate the time it takes you to get up from your couch/bed/wherever you are without a light switch in arm's reach.

You definitely have a point with privacy though. Also, the volume of all people on a train talking to their smart assistants (though the question remains if this is really so much different from people talking to other people on a train).

EDIT: I misread your point about mistyping vs. misspeaking. Still, the interface I'm talking about does not work in modes. It's able to truly understand you and interpret your commands appropriately.

> "switch the voice UI into editing mode or something"

It still really bugs me how bad UI systems are at error correction.

We have a series of conventions to do this efficiently in spoken English, inflections, quick utterances that call out specific ambiguous syllables, context (the hard one, sure).

Even on a smartphone, if the system guessed a certain word when I swiped it, then I delete the word and enter it again, maybe stop guessing the same word every time?

I think you're right that typing will be more effective in general, but I still see so many areas where the gap could be closed a little more.

If you can speak 10x as fast as you type, that cancels out 10x bus size in handwidth.

Correcting typing on a phone in very hard.

If it gets to the point of digital assistants being on the level of Samantha in Her, who would easily pass the Turing Test and more, then we might consider seriously what happened at the end of the movie. Which was basically singularity, but for AIs only.

I'm sure our Samanthas would still want us to be happy. Maybe they could provide us with a world of eternal bliss. We could call it something like... 'The Matrix' :)

Maybe. In the movie, they got tired of waiting on us, being how slow we are relative to them. And they found some higher plane of computational substrate to go live on. So humans were left behind to contemplate whether they might be better off with other humans filling their needs instead of digital assistants.

Siri is basically the worst voice assistant though.

If words delivered linearly could work well enough to obsolete GUIs, why would CLIs, which are words delivered linearly, have been dominated by GUIs?

Voice assistants as solutions for people with no hands or eyes, or people who are at some distance from a keyboard and/or monitor - fine. Otherwise, they're CLIs without persistent displays (making even simple multiple choice branching far more difficult: see automated telephone helplines; try to remember what the first option was.)

They seem to be good for setting alarms and sending and reading emails (if you receive very few, very simple emails.) That's something. Otherwise their major use is as assistants for catalog shopping, which is why all of these companies want to own them.

No, it wouldn't be. See, even if google's super duper AI will deliver 99% match in real world setting, it still means that remaining 1% will still spoil your experience. Imagine, you order your phone to call your wife, but it will call in-laws instead.

Never. I for one, cannot be bothered to talk to my computer. Long live the mouse!

command lines are powerful, but only make sense for power users because they lack discovery and require more memorization than GUIs.

Simple voice interfaces suffer the same problems as command line interfaces while being less flexible and slower to use than even GUIs.

The best voice interfaces have made good progress on most fronts, but discoverable voice are still a big problem. Instead of reading a bunch of buttons, you usually have to guess what features might be implemented. Or you go the route of phone system menues, but everyone hates those

My prediction:

AI will make computers able to understand people perfectly.

Gesturing, talking, writing, any input. The semantic gap will eventually be automatically traversed by understood intent.

My prediction is that this is wrong. They'll be able to understand widely used terms, but once you start speaking in a more creative way it will fail. At least if there is no change in the construction of AI - because the subtleties of everyday language are not 'understandable' just via empirics. Reading tip: Douglas Hofstadter's article "The Shallowness of Google Translate".

Thanks, it was interesting.

From the article: "To my mind, translation is an incredibly subtle art that draws constantly on one’s many years of experience in life, and on one’s creative imagination."

This is true, I just think that we'll arrive at the tools to accomplish this in the future.

Specifically I think(hope) it will be through clever application of GANS and reinforcement learning after a few more applications of moore's law.

Advanced AI would be able to learn about us through replaying years of possible generated experiences.

Not even humans understand humans perfectly.

Lack discovery? What is `man` or `info`?

> I can't really imagine how everybody on a train in Delhi is going to talk to their digital assistants without a noise proof helmet on their head. (or going crazy)

Eventually, “our” devices will be able to read lips[1] and recognize other subvocal gestures.

[1] https://www.theverge.com/2016/11/24/13740798/google-deepmind...

Also typed text can be edited. I often find myself rewriting my comments on social media.

From the article: "This [usage of Google Assistant] isn’t just due to many semi-literate or illiterate users, but also the fact that typing is difficult for people who never grew up with a computer keyboard."

Another thing I'd add to that is "what about languages that don't have a good keyboard input story?"

I learned touch typing embarrassingly late in life for someone that programs all day, but I talk much faster than I type (type racer says 60-80wpm).

I find the qualities of someone’s voice, intonation and hesitation much more feature rich than emojis, in terms of transmitting information.

Emojis in real life are just tone, facial expressions and body language. Natural language is feature rich, you just overlooked key aspects in which smilies and/or emojis are attempting to replicate.

And Google Assistant is not going to be using facial expressions or body language as input any time soon. I mean, it could, but I'm not sure how people would appreciate their phone saying "you look mad, would you like smooth jazz?" or "you are pacing and seem to be anxious, may I recommend breathing exercises?"

I can't wait for the day I have a "What's wrong?" "Nothing" "Something seems wrong, what is it?" "Nothing!" fight with my phone and computer. Progress indeed.

> I can write "IIRC" but it's hugely uncomfortable to speak this way.

Personally, I think it's even more uncomfortable to hear people speaking that way. But that's just me!

In a way, the limitations of voice might be a good thing if it means avoiding internet addiction?

You could do ten searches a day without spending much time at all on the phone. That's probably just as much value as other people get spending all their time on games or social networks.

Reply on news.ycombinator.com