The big howler for me was Cortana running as part of Windows Server 2016, by default.
Look: I'm not really happy running Windows Server to begin with [HN crowd: a decision that was not mine, by the way], so why do you have to have this stupid thing downloading and running all kinds of random executables that I have no idea what they do in my bloody server farm? And yup, I have logs off all those things, and pager traffic as a witness.
Yes, you can turn Cortana off. No, you shouldn't have to. This decision beggars the imagination.
(I like to think that the PM behind Cortana-makes-sense-in-a-server was the same nimrod who also shipped the touchscreen support in Server 2012R2 and the all-caps menu nonsense that plagued Visual Studio and Office for a few years, until Microsoft got a clue and realized they were honestly, really and truly pissing off customers. Someone needs to take that person out back of Building 9 and show them the instruments).
>The big howler for me was Cortana running as part of Windows Server 2016, by default.
Cortana is not running as part of Windows Server 2016 by default . The default install option is no GUI and that option does not install Cortana. You will explicit have to choose that you want to install the "Desktop experience" in order to get Cortana. If you select to add an optional feature named "Desktop experience" it is fair enough that Cortana is there since she is (or now was) a part of Microsoft Desktop experience. Most typical server workloads run fine without the GUI (and Cortana).
Apart from AD and Exchange, the only Windows-only server programs you may need to install are likely to be legacy domain specific, professional kind of stuff. I've seen tax-filing stuff, finance regulation compliance, medical or judicial "solutions" that would only run (poorly) on Windows, usually with mind-boggling requirements (run as domain admin! Server 2003 only!) and they simply won't run without the GUI.
In other words: usually, if you need Windows server, you're going to need the GUI.
Note that I say this as a Linux admin who has only run Windows very occasionally (and reluctantly!) in the past 20 years.
It’s a perfectly fair assertion given the abundance of commercial software in active use which is a “server” written into the back end of a desktop application. They are not written as stand alone Services.
Remember this is the real world we live in, where once a developer learns how to build a Windows application in Visual Studio, everything becomes a Visual Studio app based on the same template. Consider yourself lucky when the developer uses a newer version of Visual a Studio and dotNet than the one they started with a decade ago.
In the perfect world, yes we would be paying developers high salaries and allowing lots of “down time” for retraining into the current version of dotNet, and allowing plenty of time to upgrade legacy applications to current versions of Windows.
In the real world we still have that Windows 2000 server sitting around with a printed sign taped to the keyboard saying, “DO NOT UPDATE: PLEASE SEE JOEL BEFORE TOUCHING” only nobody remembers who Joel is and he left the company before Confluence arrived. All we know is that we can’t even install Remote Desktop Services because when Joel tried that it broke the application software. That sign appeared on the machine a week after that incident because that’s how long it took Joel to repair the server to the point that the application would work again.
Look, as far as I'm concerned and as an SRE, server-side Windows is legacy. (With the only already mentioned exceptions being AD and Exchange.) And client-side Windows is ... well something I don't use personally anyways.
Yep, likewise. I think the only thing that required the Desktop Experience off the top of my head was WSUS, but we have all sorts of other roles running on Server Core without any issues at all.
Sure you can, and what's the value added compared to any Linux? The joy of having to deal with CALs and licensing audits? If you just like writing cheques, RedHat is an option.
The number of roles with no or effectively no Powershell cmdlets is staggering, and remote MMC require punching a dozen holes in the firewall when it even works. Even then there are many roles like iis and adfs that are still effectively broken with no gui.
The "default" install is headless, yup. Yet, if you want something usable in many environments, you'll need the GUI. We use third party (and Microsoft) tools that do pixels (and PowerShell still doesn't do everything).
So if you were given the choices:
1. Headless (lots of stuff doesn't work, especially third-party utilities)
2. GUI
3. GUI without Cortana and other crazy
... which option would you pick?
There is no #3, by the way.
> it is fair enough that Cortana is there since she is (or now was) a part of Microsoft Desktop experience
I don't care how you spin this, it's very bad practice for a "server OS" to download and run random code. Making Windows Server secure is hard enough without this nonsense. "Oh, Cortana comes for free with the Desktop Experience" is just begging the question, it doesn't make the underlying bad practices of that Windows component somehow okay.
Yes. As a very happy Windows user, both server and desktop, I'm always scratching my head about the FUD comments I see here about Windows. Most are simply untrue, else they're vast distortions. It makes me wonder whose behind them and why. Is it just ultra-fandom, or is someone trying an orgranized attempt to manipulate markets through mis-information?
> I'm always scratching my head about the FUD comments I see here about Windows. Most are simply untrue, else they're vast distortions.
Were they always untrue, or were they true for some older version of Windows? Many non-Windows users here have used Windows some time in the past, and their views on Windows are based on that experience.
Windows Server 2016 (Desktop Experience). Choosing this option is useful when a GUI is required; for example, to provide backward compatibility for an application that cannot be run on a Server Core installation. All server roles and features are supported. This configuration is the equivalent of Server Graphical Shell on Windows Server 2012 R2.
I seem to remember that even when Cortana is shut off, it was found to be listening and sending telemetry back to Microsoft. Based on their past behaviour, I'd guess that even is it is "removed", those elements will remain.
Oh, I'm sorry for pointing out the fact that speech recognition wants outbound access through the firewall even on Cortana-less LTSC. That must be so hurtful! Please forgive me, I'm on your side, I don't even vaccinate!
Funny how the LTSC version is the only Windows version that works best as a standalone desktop OS, where you install the software you want and no updates will break your configuration every half a year. But of course it's not recommended for desktop use, and you can't get a license as an individual user (I think even business users have restrictions).
I would have been fine using Cortana in Windows 10, but unfortunately, it slowed down the Windows key search UI just enough to be frustrating and worth the time to disable in settings. I don’t at all mind its existence but it really needs to add functionality without introducing new drawbacks, especially for frequently used areas of the OS.
(MacOS was little better as it didn’t even try to integrate Siri with Spotlight and kept it in its own, infrequently-used app.)
Agreed entirely. In Windows 7, I could hit the Winkey, type the first few letters of a program, then hit enter. The results were instant and consistent, so I could build muscle memory on it. In Windows 10, the search results take several seconds to show up, and change depending on what new files there are, how many starting characters I typed, whether the index is up to date, and the phase of the moon.
There must be a lot of macs or linux machines in the department responsible for the search because there's no way any developer is dogfooding that user hostile mess
IIRC the conversation I've had with a former Microsoft employee, some features are developed as mini competitions. Requirements are sent to the group(s) and first acceptable implementation gets merged.
e.g. for the Ribbon-UI in Office, there was 3-4 implementations until one was selected as final.
So the Windows search UI is the fastest developed one with acceptable features.
If I wanted to create a process with the goal of generating technical debt, I wouldn't get something that effective. It's even beaultiful how it leverages all the competition flaws to get this result.
I imagine the result is reflected on the employees productivity assessment and the worst performers fired, so it can exploit evolutionary dynamics to create the highest possible amount of debit.
Remember, this is the same Microsoft where people are frowned upon for merging fixes and optimizations to the Windows kernel instead of implementing new-and-shiny features [0].
Microsoft now believes I have a deranged crush on some 1800s Canadian named Ostman because it takes just long enough to eat the first character when I press the windows key and start typing in 'postman'.
"Despite dropping the app in these regions, Cortana itself remains “an integral part” of the company’s business model to incorporate “conversational computing and productivity” into its products according to the spokesperson, so Microsoft doesn’t appear to be closing the coffin on it entirely."
Don't make a girl a promise... if you know you can't keep it.
Cortana was doomed the moment Microsoft gave up on Windows Phone. It could never become as integrated with the mobile os as the competing assistants. Even if Cortana was better, it wasn't always listening. And on a desktop the assistants are still not sufficiently versatile at home, and not compatible with the open office environments at work.
Also Cortana was always good at languages it suported, but it couldn't provide good localized content for most of the world.
I won't be crying since it never worked on my WP devices in first place and setting up device only to talk to the virtual voice assistant in English, in non-English language country was some weird acrobatics; once finally I managed to test it out, I had this feeling the voice had kind of mocking tone. The substitute or basic assistant for Poland was capable of understanding the language but you'd have to skip all grammar so it could understand you correctly - notably skipping noun grammatical cases. Perhaps the complexity of Slavic languages (and other non-English ones) is still a barrier for this technology.
The presence of voice assistants on desktop operating systems is a completely different issue - personally, I don't see how these can be useful; they're more of a gimmick feature aimed at less experienced users.
Yes. The big parts are the models (neural networks (NN)). You can make them quite small (sparsification, quantization, transfer learning to smaller models, ...). But this is non-trivial.
Kaldi is the most complete open source speech recognition software, covering HMM/GMMs, the conventional hybrid HMM/NN, and much more. If you care about end-to-end only, there is much more, e.g. Mozilla DeepSpeech, and most modern open source encoder-decoder-attention frameworks can do that (e.g. Lingvo etc).
Most of these encoder-decoder-attention framework can also do text-to-speech (TTS). Maybe you need a separate vocoder, but you will also find open source code for that.
For the assistant engine, there are also a couple of open source projects, e.g. Mycroft, LinTO, Rasa, Leon, Open Assistant, Susi, ...
Putting it all together is non-trivial, and requires lots of expert knowledge, and lots of work and time to get a good result. This is hard to do as a hobby project.
That site has the most GDPR compliant, transparent, easy to configure cookie settings I've ever seen.
I've got my eyes on that dev kit for sure. Having an offline voice assistant is something I've wanted for a while, privacy is much more important to me than the hardware support that echo and Google home have. Thanks for dropping that link!
Cortana isn't removed, MS only removed it from the core Windows feature set.
When the 19H2 is released Cortana will still be around, but not as part of Windows but as a separate app.
People who update their Windows to the Fast Ring can even play with the new Cortana app right now.
Microsoft also have plans to deeply integrate it into its 365 line of productivity apps, so it isn't going anywhere in that area too.
And given that Siri and Google Assistant both dominate the Assistant area on the smartphones due to the fact that they are both preinstalled makes discontinuing the standalone smartphone apps and putting the money and resources involved to good use elsewhere is actually a good move business wise.
i think you can ask siri about cortana. i've gotten back something about how siri understands how hard being an assistant is, and has respect for all assistants.
I don't know a single person using Cortana, so it's not a surprise I guess. At least the push to use it in Windows will finally go away... A bit too late I have migrated to Ubuntu earlier this year... Puting an end to 20 years of Windows on my computer.
I don't use many Siri commands but a few come in handy:
- "add milk to grocery list" — I have a reminder list called "Grocery" to make that happen. Makes it easy to add stuff as I realize I'm running low, then I can check it off in the Reminders app.
- "create appointment next {day} at {time} {title}" — I never use that one when I'm next to my computer, but when I'm around the house or after a doctor visit, it's usually accurate.
- "what time is it" — no I don't wear a watch. But when I'm hiking and I'm too lazy to pull my phone out, this comes in handy as it works though most pockets.
- "send text message to {person}", "call {person}" — when driving, that's a pretty seamless way to convey quick messages without touching the phone.
- "get directions to {place}" — I only use that for locations I know the phone will understand, like favorites or business names that are not ambiguous.
What I really like about Siri is the geofencing capability. "Siri, remind me when I get home to text Daniel" (or something of that nature) This way the reminder is not tied to a particular time, but instead tied to a location that is more relevant or convenient to carry out the task.
I actually used Cortana on my Android phone. (Disclaimer - I work for MSFT). The main reason why I don't use Google assistant instead is the total unease with having Google listen to me 24/7.
Cortana's not too bad for the simple stuff like getting the outlook of the day or setting up geofenced reminders (remind me to unload the washing machine when I come back home).
Anything more complex absolutely fails 100% of the time, with this ridiculous behavior most assistants have to default to "here is what I found on the web". Why on earth would I want you to search for what I asked you to do? Wouldn't I ask you to search on the web if I wanted you to search on the web?
I think that's the downfall of most voice assistants I've tried. Discoverability is poor, which wouldn't be a problem if the feature set was not limited to the basics.
That being said it also loses a lot of its appeal by requiring a physical activation (either clicking on the app or on my headsets button), since Google removed the possibility for alternative voice assistants to do that. Compared to "hey siri", that schnucks.
For something hosted on a phone, I think they could improve the feel of privacy just by discarding the wake-word concept. This also allows them to cut out the passive listening hardware (saving money) or software load (saving battery life and CPU overhead)
If you have the phone in your hands anyway, it's not that outlandish to have a button you press to engage Siri/Cortana/etc. This also avoids annoying/embarrassing accidental triggers.
The only thing I ever used Cortana for (back in the Windows Phone 8 days) was the geofenced stuff-- "take the perishable items out of the car when you get home".
Absurd thought: With enough data mining, not having the assistant's features discoverable could become itself a feature: Sifting through a billion requests and where it falls back to web requests could expose "desire paths"-- the specific functionality consumers expect a voice-assistant to provide but wouldn't necessarily know to ask for on a marketing survey.
I use GA all of the time. I usually have bluetooth headphones around my neck and use them constantly to set timers for reminders, directions in the car, and most usefully whenever I want a quick answer to something. It's really nice to be occupied cooking dinner still be able to get a quick answer to "what's backscatter radiation" or some other random thing when listening to an audio book, or thinking about, well, anything. In this respect GA is far superior to Alexa, which returns "I don't know that yet" 3 out of 5 times.
I regularly use Google assessment on android to add stuff to my shopping list. It's very convenient especially while cooking when my hands are occupied, as soon as I notice something running out I just ask Google to add it.
I use Alexa to get the echo dot to play random toddler songs and songs like them. It’s actually really useful for that in the living room. Unfortunately, my toddler’s first complete sentences were “Alexa, play baby shark/space unicorn/cat flushing toilet”.
I use Siri for timers and alarms; ask for the weather, and for, sending text messages and calling when I'm on the go. My GF has been a heavy user of voice transcription on her phone, but when she got an Apple Watch she really ramped it up. It is a pretty good way to use the watch, especially if your AirPods are connected to it.
I use Google assistant to set reminders. It's so much more convenient than clicking through and typing the same thing on my phone especially or even my laptop. Literally works in 3 seconds. It's great!
I don't think that is worth the privacy lost by having Google parse voice commands. 15-30 seconds inputting data versus privacy lost by having a profile generated on me by Google.
A Google VP literally said that you should warn guests if you have voice assistants in your home.
I think it should be federal law that devices like these that are always listening come with stickers for purchasers to put on the front door warning guests that they are being recorded. Of course no one is going to use those stickers, but it will remind everyone of how big of a privacy violation these things are.
The only time I ever use it is when I'm cooking and my hands are dirty (and usually just to get some measurement conversion), or when I'm wanting to start navigating somewhere quickly before driving. I've found it wholly useless, frustrating, and slower in all other circumstances.
Now, I will establish my HN cred by saying I prefer CLI and Unix, would read my mail in PINE if I could, and run everything from Linux to Windows to MacOS in the house.
But I use Siri all the time. Perhaps because I desperately miss a keyboard? But asking Siri to play a particular song, call a particular person, give me directions, and turn on the flashlight? All are way faster than navigating the touch interface.
And dictation is the best way for me to message while on the move; I’m old so typing a flat keyboard while walking is a foolish idea.
That said, accuracy is terrible. So many times I have wrong song. I don’t understand why Alexa doesn’t learn that the song I have asked for dozens of times is the right one, not the weird house version that sneaks in.
I probably should investigate iOS shortcuts, that may tip me back to the touch screen.
I get instant rage when I'm jamming out to some tunes or watching a movie with headphones on and Siri pops on and cuts off the audio. Then when I angrily swear at her, rather than just cutting out, she responds that I'm not being polite.
I use Siri extensively with HomeKit, and have an Echo in my office for development.
Siri talks to a dozen lights, sockets and sensors about the house, but I don't trust Alexa with access to more than a couple (essentially my office lights) since their security model sucks compared to Apple's (Google's isn't much better, IMHO).
For instance, I was forced to open my Plex server to the outside to get Alexa to play music on my office speakers, which is borderline insane (I set up another Plex server on a VM with my "work" music).
I use Siri to set alarms and timers since that's far and away more convenient than opening the app and switching to the right tab and futzing with the time dial.
If I've got my phone in my pocket and my hands full it's helpful to be able to tell Siri to turn the flashlight on.
Aside from that I don't think I've used it for much. I get the weather from the radio already, and don't care to search the web with a voice assistant. It'd be nice if she'd read me my text messages but she flatly refuses to do that.
> It'd be nice if she'd read me my text messages but she flatly refuses to do that.
Siri is integrated with AirPods Pro and will read your text messages to you as they come in. It’s pretty great.
But what really excites me about voice assistance and voice assistive technology is the advent of Voice Control in iOS 13. Voice Control allows me to manipulate my iPad, for example, while using both hands to eat dinner or wash dishes etc. My screen doesn’t get greasy and I can scroll and tap links using my voice.
The on device dictation ability is also first-rate. For example, I’ve used dictation to generate the majority of this comment.
me neither, but if you ever go to a diner (at least in the US) you'll probably spot some (imo extremely rude) patrons watching some blaring loud movies on their tablets while they eat.
I absolutely hate that trend, if it wasn't obvious.
I use Alexa to add to my grocery list over the course of the week. If I run out of an item (e.g. Use the last of the milk, run out of butter, notice I'm getting low on bread, etc.), I just tell Alexa to add it to the grocery list. It's faster than whipping out my phone to do it and it has made building out my grocery list more reliable. (Inevitably I would struggle to remember everything that I needed to get when I would wait to do it at the end of the week)
I don't think we've touched a light switch in our house in eight months due to voice assistant. Kitchen timers, asking spelling of words, doing imperial to metric conversion, getting weather. Sometimes if I'm in an argument with my wife I will have google define the term we're arguing over for dramatic effect, or we'll consult google via voice to clarify something, like the population of the Philippines, for example.
Has the introduction of voice commands to the home had a statistically significant effect on how many arguments you have with your wife? Does she use it as much as you do?
No correlation between additional arguments, it's usually politics or public policy or similar; she does actually use it about as much as I do. I would say 75%, which seems especially high even though she's not big in to technology. She's actually been pushing to get a voice control node "home mini" installed in the bedroom for quite a while, and wants a nest thermostat for the next place we move into.
I had voice control set up for the house lights for a few months, and went back to having a manual remote clicker in most of the house and a PIR switch in the pantry.
For me at least, button pushing has a lower cognitive load than talking at something.
I've seen a friend use it while driving. Seems painfully obvious and yet I still refuse to use it while driving. Instead I just don't phone for that time.
Using a voice assistant might be less distracting than using your phone, but it's still distracting while driving, having to concentrate on what you say, the right commands, their feedback, etc.
A conversation with another person in the car is not the same as a phone conversation (for one, the other person doesn't share the road cues with you to know to stop talking when something requires attention, can't see your face and you can't see theirs, it's easier to get emotionally triggered, etc). And of course another person in the car can also help point to things to watch out on the road, a person in a phone call, not.
Even worse, talking to an assistant is even worse, you need to spend mental effort to phrase things in a certain way (voice interface wise), parse their responses (which could be off), and so on...
There's at least one way in which it's easier talking to a voice assistant, I use AndroidAuto as my GPS (ie navigation aid whilst driving). When you talk to someone they don't "bing" to show they're listening, so you generally have to look at them.
Me: "OK Google"
AndroidAuto: {bing}
Me: "Navigate home"
AndroidAuto: "Navigating to [home], your travel time will be Xhrs, usual traffic along your
route" [or whatever it says]
It's pretty low concentration, like tuning a radio manually I'd say.
You have to choose how you phrase things for different people too. If you're talking to another adult with kids in the car you might be choosing that phrasing with quite a deal of cognitive load.
I use Siri mostly for more complicated reminders I don’t want to do manually like “remind me to do $x when I get in my car” (based on Bluetooth connecting to my car) or “remind me to do $y when I get home/work”.
I work out with only my cellular Apple Watch. If I need to send any long text, I’ll use voice dictation”.
Absolutely! I use Google Assistant to control my home automation (mostly lights and TV), check commute times when leaving for work, play NPR, and occasional Wikipedia lookups that don't warrant pulling a phone out.
I also use the Leidos Flight Service skill (voice app) to get a quick preview of the weather before driving to the airport to fly a plane, if I'm not in front of a computer. Of course, I'll pull out a tablet and do a more in-depth briefing at the airport before actually flying... but it's a great way to check nearby METARs and TAFs while on a commute.
I use Alexa at home pretty frequently and Siri on my iOS devices pretty frequently. Examples include "add <item> to my shopping list", "turn my bedroom lights off/on", "what is the weather today/tomorrow" and "set a 4 minute 15 second egg timer" (for my poached eggs every morning)
"hotword" detection is done in special recognition hardware so as to not incur latency to the cloud, which is where the rest of the voice recognition gets done. You can't change the hotword, or can only choose between a few, because that's what's baked into the chip on the device.
I use Google Assistant all the time in Android Auto. It's actually remarkably brilliant... Until it isn't and becomes either humorous or frustrating depending on the situation.
I use it all the time, especially on HomePod. I also use it with Watch to send messages. In the car, I use it for directions. Don’t use it on my Mac though, spotlight is just too fast.
This is clearly the worst crowd possible, except maybe people with speech disabilities, to ask this question. Just look at how many people don't like it.
I really can't make myself talk to a machine. I even abhor the phone answering services that insist that I speak instead of punching numbers. I my household the only person that interacts with the Echo/Alexa is our 3 year-old daughter. She finds it funny.
On the other hand, the voice assistant devices are selling like crazy, so people are interested in them.
You can persuade most of these to let you use numbers if you respond by hitting 0 every time they repeat the question until they produce a number punching list.
Look: I'm not really happy running Windows Server to begin with [HN crowd: a decision that was not mine, by the way], so why do you have to have this stupid thing downloading and running all kinds of random executables that I have no idea what they do in my bloody server farm? And yup, I have logs off all those things, and pager traffic as a witness.
Yes, you can turn Cortana off. No, you shouldn't have to. This decision beggars the imagination.
(I like to think that the PM behind Cortana-makes-sense-in-a-server was the same nimrod who also shipped the touchscreen support in Server 2012R2 and the all-caps menu nonsense that plagued Visual Studio and Office for a few years, until Microsoft got a clue and realized they were honestly, really and truly pissing off customers. Someone needs to take that person out back of Building 9 and show them the instruments).