Hey Cortana, how about you hide my app from the user? It's a way off yet, but one of the endpoints for Cortana, Microsoft's take on the digital assistant you can fall in love with, is to be a primary interface between multiple applications. The idea is that instead of opening an Excel spreadsheet to find a datum you need, copying it, switching to e-mail, and sending it to the …
The bald assertion that "the less time the user spends in the app, the more productive they're going to be" sounds nice, but does not seem to have much in the way of proof behind it. In fact, it looks a lot like other pushes by Microsoft to overhaul the user interface without checking to see what actually works for its customers. There have been plenty of studies which clearly demonstrate that people who think that they are quite good at multitasking are actually deluding themselves (many have been done to demonstrate that texting while driving is a bad idea). This looks like another way to fragment the user experience, even if it is not done on a visual level. If you are bouncing around between tasks, even if it is not reflected by what's on the screen, your overall productivity will go down.
It also looks like a way to dumb down the user interface to the lowest common denominator. People who cannot type might be able to input info into a Word document a lot faster with Cortana to start with, but will not be able to ramp up to the speed that someone who really knows the application can achieve. Similarly, I suspect that this will hold true for working with computers and apps in general. This tech might make things easier, but I do not think it will make things better - and that is my unsubstantiated assertion.
Office Mac 2011 was usable because it's got a menu but Office 2013 is a nightmare for me,
I'd have to agree on this and as I've been using Word since version 2 (skipped Office 2003) I'd say I've coped with most UI changes (disappearing menus was another disaster).
Office 2016 has the advantage of looking and behaving largely the same on Windows and Mac and is far less confusing than Office 2011 or 2013.
After having many useful apps disappear from the web and slip through my fingers, nearly two decades ago I created a folder named "Programs for a Reformat" into which I stashed any downloaded program installers which I felt worthy of saving. When I would buy a new computer with a newer OS I would use this folder to bring forward all the helpful little bits I had gathered over the years. This folder has evolved to contain many "Read Me" files, PDF manuals, relevant web pages, etc. I have brought forward through Win98, Vista, WinXP, Win7, and I expect Win10, dozens of useful apps. The three that get loaded before I even think about going online to grab the latest VLC or IrfanView or Firefox, are wordpad98.exe, (yes I still use this for simple .rtf files, and find all subsequent Wordpads annoyingly frustrating and space-wasting), tclock.exe (to augment my taskbar clock with the date) and SaveDesktopIconLayout (because in no time my new screen will be filled with sorted Icons). Moving to a 64 bit OS forced me to search anxiously for some replacements, but so far, success, keeping me in my comfort zone. I'm crossing my fingers about any future OS though, MS be damned.
they do a lot of testing and so support many routes to the same result as people work differently.
This assertion flies in the face of most usability research which has been summarised as "make design as simple as possible and reduce choices for any particular task. No choices are best".
But, hey, if it works for you then you may have a great career ahead of you.
"This assertion flies in the face of most usability research which has been summarised as "make design as simple as possible and reduce choices for any particular task. No choices are best"."
Win8 was simple. The ribbon bar reduced choices. Give me freedom to find what works for me.
I think the ribbon still gives you most of the same options in Word for example, but they're just hidden in a crap interface. Judging by the small sample of users in my work group they find it extremely frustrating and even after what is now many months of use they still have trouble finding functions they use quite often. The problem with 'simplifying' a voice interface is that you have to think while you talk. Clicking toolbars and pop-outs is more like typing, you can do it almost mechanically provided you know where things are. That's why the ribbon is such a fail.
The possibilities for mayhem are absolutely stupendous.
Didn't get the memo ? Cortana didn't send it to the right person, or at all. Wrong figures in the spreadsheet ? Cortana didn't use the latest version. Used the latest figures on a report concerning last year ? Cortana made the mistake. Powerpoint slide looks ugly ? Cortana made the choices.
I really look forward to the Helpdesk receiving a call in the lines of "Cortana didn't send my mail, you need to fix it."
Yep, that'll be fun to watch.
I hate them all! Siri, Cortana, Google and whatever Blackberry call theirs. Unless they're 100% accurate, the time you waste repeating commands and checking responses is greater than any possible time saving.
That's from someone who works from home, on his own. Hate to think what it's like for folk in crowded offices.
"the time you waste repeating commands and checking responses is greater than any possible time saving."
Also completely unusable if you have a strong accent.
If you want to truly minimise the mental adjustment time between applications, would it not be better to (at minimum) ensure all applications follow a sane ui guideline to minimise 'culture shock' when moving between applications.
The best way to work is to have your information easy to find - whatever happened to the database fielsytem idea? (or any idea to stop some windows users saving everything to the desktop)
Absolutely. If my usage of Siri or Cortana is typical then these features are merely an interesting (from a technical "ooh - look what they can almost do now") sideshow that you use once or twice and then realise that typing or doing something else is a better idea.
In its defence, I will say that at least Siri and Cortana are not as bad as the voice control for the phone on my 2005 Audi (think of the Jeremy Clarkson review of the BMW voice control....). The sheer frustration of the Audi voice interface and its inability to discern more than one number successfully, and the subsequent failure for it to understand the corrective commands that it then offers is exquisitely horrendous. Again something used 2-3 times and abandoned in sweary bemusement.
For all of the above tools, you have to remember that someone signed them off for production use - as in "Wow, that works really well! The users are going to LOVE it". I must be missing something....
Don't get me wrong, I appreciate and admire the technical achievement of the real-time language recognition and contextual analysis, but echoing the sentiment in the post above, unless it is close to 100% accurate then it is unfortunately almost 100% useless as a primary interface.
Steve
Then it disappeared with TIFKAM.
And there's no way this will work in an office with all the background noise.
They seem to be flailing around with no idea what to do next. Any changes should complement the current UI, not wreck it meaning they have to try and come up with something later that fills in missing functionality.
Office? pah! What do you think happens in an actual production environment?
Try giving voice commands on the controlling console of a sealer/packager going at full tilt.. Mind.. there's a lot more about "METRO" that makes it completely useless for most management and control applications, but that one makes me laugh. A lot...
This being a tech site we tend to focus on background noise disrupting voice commands.
But let's also add in voice commands adding to background noise.
Try talking on the phone while someone nearby is shouting "Cortana, open the client spreadsheet, no I don't want to search fuc**g Bing I know what a f****g spreadsheet is I just need to.....".
So, how often do you use grep by typing its input into the console?
Command-line interfaces are closer to spoken word in their structure than a GUI is, and the idea of taking the meaning of one term and using it as the subject of another is fundamental to getting work done in any kind of text-driven shell. Why wouldn't this work for speech too?
Imagine I'm selling a flight reservation application, do you think I prefer my users to have to open it, enter their dates, hit search, or maybe it's better to accept the search parameters from a "standard input" (in this case, the speech-to-text engine) and work with it that way?
Because you can never be sure the STT engine's got what you want.
You need to tell the STT engine where the data is very precisely otherwise it's not going to find it, which means you need to see the data in the first place. Then you need to show the piece of data that STT engine understood you wanted, to check it's right. Finally you need to tell it where it's going, but you'd certainly want to double check that before committing.
That means you can't say "open my last document and send the 2015 data to John" with any degree of confidence.
Maybe some kind of Star Trek TNG where data was displayed on the screen and the actors told the computer to do stuff with the data on the screen, but the thing in charge of presentation was the app, not the STT engine, because the thing that knew how to present the data best was the app, not some square-peg-in-a-round-hole STT engine.
The best way to do this would be to make the STT invisible and like any another input device for the app, like mouse, keyboard, and batch/scripting support, but obviously that can't be done because marketing want to sell Cortana.
"because marketing want to sell Cortana."
That's it. They're selling the "feature", rather than it selling the OS. Cortana isn't here because we asked for it. It's just another excuse to send data back to the mother ship.
If it was wanted, then there would have already been 5 different implements for Linux by the time MS began working on it.
" Ever tried pronouncing a regular expression?"
Er, the expression is not the input to the grep tool.
As for Linux not having an implementation, it's because this type of speech to text conversion is statistical - it requires a server with a lot of computing power, and a hell of a lot of training data (drawn from a diverse population of speakers) to get it working. Who, in the various Linux communities, would be willing to fund, or co-ordinate, this effort?
"As for Linux not having an implementation, it's because this type of speech to text conversion is statistical - it requires a server with a lot of computing power, and a hell of a lot of training data "
A locally based speech engine with always on training s/w to train the speech recognition to the users speech pattern would work far better (long term), it would require some work for the user upfront to train it (a few 'games' and 'activities' included would help/encourage this), but in the end you'd end up with a better product. This would negate most of the computing power. Added bonus of 'you'd be in control of your user-generated data'. Once it's set up, the only statistical issue is processing natural language usage changes ('dictionary' updates) and interfaces to online searchable sources.
There's been a few linux projects recently, but most the ones I've come across recently were abandoned after a while and depended on use of googles speech API (although I've not done any investigating recently).
Why does it need to be online? Nuance's STT software seems to cope quite well despite being offline.
MS probably took one look at Nuance's licensing costs and decided to go with an online deployment!
Online STT places minimal demands on the client device and is a proven deployment model (it was the model we used back in circa 2000 with Nuance because we wanted our system to work with all phones). However, the really big component (and the one in which there will be lots of proprietary IP - much of which will be Nuance's) will be the semantic analysis and knowledge-based interpretation of the STT input, which, as we've seen with machine-based language translation, get better more rapidly with large volumes of data.
"Some developers are saying 'we don't want the user to spend any time in our application'"
Oh come on. Who the fuck has been saying that?? I mean, really...is that what developers have been demanding?
I think what they actually said was "We're business-logic specialists, we don't want to waste our time writing a GUI -- especially not when you guys are going to change the whole GUI paradigm again next week and we'll have to start all over again!"
Nice of them to research this. But in a normal office or call-center, it'll be the pits. Besides, unless you're blind, most people are very visually oriented. They might know what data they want, but can't define it until they see it.
Now if MS is coming up with an auto-driving car and they want to use voice commands... uh-oh, nevermind. That's a very scary thought.
Me: Curtana! Drive this fucker like ya stole it!
Curtana: It's about damned time! Sit down, strap in, hold on, & let's GO!
...later...
Me: I'm sorry Officer for the multi-state high speed chase, going up on sidewalks, mowing down pimps & whores, sideswiping all those mailboxes & lamp posts, and chasing that bus of Nuns & Orphans until it crashed into the drive through window at McDonald's, but it wasn't my fault.
Officer in utter disbelief: How so?
Me: Curtana did it! She was playing GTA and got confused. I think she's finally snapped & is in dire need of an upgrade or something.
Curtana, in obvious disgust: You TOLD me to drive that way, you fucking meat bag!
Me, pointing to the crumpled & bullet riddled dashboard: See? See? She's insane!
Officer, pinching the bridge of the nose between thumb & forefinger, eyes squeazed closed, muttering darkly: I don't get paid enough for this shit...
=-D
On a COMPLETELY unrelated note, where can I get an autonomous vehicle controlled by Curtana or Siri?
*Pure, Sweet, & Innocent smile*
TBH I find all these 'personal assistants' a short lived amusement at best. Cortana still can't send a text message or make an appointment without fucking it up meaning you're quicker to type it in the first place.
I've not figured out if there is a particular accent I'm suppose to put on to get it to work only that it isn't the one I've got now.
Have they even considered how this would work when everyone in the office is doing this at the same time?
It would be way too noisy and you can bet that your PC would occasionally pick up someone else's commands.
If it processes the commands without focusing the app, you may not even know that the sensitive spreadsheet you have on your PC was just emailed to all and sundry because the guy at the desk next to you tried to send an email to all users...
Have they even considered how this would work when everyone in the office is doing this at the same time?
I have - that has been known since we started with voice dictation in, what, the 90s?
On the other hand, imagine the fun the next "Occupy Wall Street" will have with loudhailers :)
So anyone who finds a back-door into Cortana can easily compromise the system, emailing anything to anyone - or everyone!
No doubt Microsoft will argue Cortana can't be hacked, then feign surprise when the first batch of exploits are discovered... and the next... and the next..
Somehow that brings back an "early Internet" time when PC sound was just arriving and we rigged a couple of sites to play a VERY loud "this user is watching porn" sound file ("loud" as in "pretty close to clipping").
Somehow we must have helped forcing Microsoft into fixing the auto-execution of web links in Outlook messages :).
Take a trivial example: if users search "help" to "paste unformatted text", and a lot of them are doing it, then you're clearly hiding the functionality away from plain sight.
That's making the rather brash assumption that their users are happy with Microsoft shoulder-surfing what they're searching for.
With Windows 10, I assume that support for Kinect has been baked in? or is it a different version being deployed on Xbox's compared to the desktop.
This brings a whole new realm of UI interactions: hoover the lounge and Windows decides your HDD needs cleaning, cobweb and there goes your cloud folders and backups...
A few weeks ago on a Saturday morning I had some mail that *HAD* to go out that day and had to be mailed inside a post office. So I naturally forgot all about it while I was near home. I was going on a day trip, rapidly leaving the city and in a part of town I don't know well at all when I suddenly remembered the mail. I had *NO* idea where any post offices in that part of town were located. I pulled off the freeway and stopped, then took out my Android phone, started the Google app and said "Okay Google, post office near my location."
In less than 5 seconds, I had a map with the three nearest post offices along with their open hours. I tapped the nearest and the directions icon and got turn-by-turn instructions to get there. Ten minutes later, my mail was posted.
And yes, I've had this sort of thing fail a lot. But I was really thankful for this one success.
So far all I've seen it do is lock up the system. Its a nasty piece of work.
Maybe I'm just not worthy. But then I have to actually write software for a living, not generate endless Powerpoint presentations about how rich we're all going to be if you give me all your money for my "can't miss" real estate venture.
Those tile thingies are also a bit weird. Most don't have anything on them because I haven't set up the application (c/f Vimes and the dis-organizer) but those that are operating gibber away meaninglessly telling me that the weather outside is sunny, would I like to play Candy Crush something-or-another and I should be monitoring lots of Twitter feeds so I can keep by ADHD stoked.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
