Comment by JadoJodo | Hacker News Reader

Comment by JadoJodo | Hacker News Reader

> is it a feature focused on search and retrieval, or an assistant that carries out complex tasks?

I'd settle for an assistant that carries out simple tasks.

I don't expect much from my robot assistant, but the number of times throughout the day that she gets it blatantly wrong has become a running joke in our household.

A few recent examples:

    Me, holding my Apple Watch near my face: "Hey, Siri, drive to Kroger <there's one I go to all the time that is 10 minutes from my house>"

    Watch: "Now playing <some band I've never heard of, that isn't in my library>"

    Simultaneously, iPhone: "Getting directions to Kroger <shows results that are ~700 miles away>"

Or:

    Me, to my iPhone: "Hey, Siri, start a yoga workout."

    iPhone: <starts playing a Yoga video on Fitness+, instead of starting the plain Yoga tracking workout I do every day>

Edit:

I just remembered another one. I was mowing the lawn on Sunday and wanted to start an outdoor walk. The mower was off, and I had my AirPods in:

    Me: "Siri, start an outdoor walk"

    iPhone: "Ok, I've cancelled your alarm for tomorrow"

I had this experience over the weekend:

Uses Siri on Apple Watch to start a timer Uses Siri on Apple Watch to start a timer "Hey Siri, set a timer for 3 minutes" "Sorry, you don't have the Timer app installed"

This gets to a problem I’ve had with Siri once it rolled out to multiple devices. Each device has a different set of capabilities. Sometimes this makes sense (eg I don’t expect my TV to be able to place a call. It doesn’t have a microphone, nor a camera.), but many times it’s just bizarre. For instance, I can’t set an alarm on my laptop. It has a clock, yet I can’t do clock things.

It seems like some basic feature (eg clock like features) should just be built in, and where that’s not feasible, perhaps some sort of device handoff, which essentially happens on the watch.

When I'm a cynical, I'm thinking bad search and bad implementations (to me) are really good implementations, the difference is that the goal is not my goal.

Examples: * when starting to type, Chrome address bar preferring to go to Google search, instead of the friggin site I've gone to a hundred times already! * Google maps, search seemingly doesn't deterministically remember my searches. If I start typing the same search term I did yesterday, well, odds are I'm looking it up again! Show that. * Windows start menu... oh don't get me started on that. :)

I'm having trouble getting Siri to even understand me. I have a HomeKit scene called "Apartment lights off" (it turns off all the lights in the apartment surprisingly enough) and I often ask Siri to apply that scene. The amount of times Siri completely misunderstands me for something strange baffles me.

"Upon lights off – Sorry, I don't understand."

"Obama lights off – Sorry, still don't get it."

Sometimes it just picks up the "lights off" part, prompting me "Which room?" And when I say "Apartment lights off" again, Siri will then reply with "Sorry, I can't find that scene in the current home." At that point, I need to close the Siri prompt and do it again. If Siri understand me that time, then I will actually get that scene applied.

"All lights to max"

"Which phone number for Max?"

I wish I could train myself to say a different phrase.

Try using groups instead of scenes. Also, key by room, or even use blanket universals (rather than say apartment).

“Turn on all living room lights” works. Also, “turn off all lights” will flip off everything in our house.

Groups are awesome. I have 3 blinds in the north of living room, so I called the, north west, north center, and north east. But they are in the north group, so I usually put them all up at once by saying “open north blinds.”

The problem with "turn off all lights" is that if I have some third-party devices in HomeKit that are not actually lights but are registered like lights, they go off as well.

Also, it's a studio apartment, so there's not really "rooms" strictly speaking.

You can always assign them to rooms even if you just have one room. But I think groups would work for your case. Like say put all your actual lights in the actual group. And then say “turn on actual lights.”

Scenes..I can’t really figure them out. They seem way too complex for what I want to do, which is to turn things off and on.

Grouping won't work at all since if I group the accessories, I will lose control of the individual lights.

Really, this is what scenes are designed for but Siri is just too stupid to understand what they are.

Scenes are activated, they don’t have direction, they don’t work at all for turning things in and off, instead you have to create separate on and off scenes.

HomeKit should really support overlapping groups, it is a feature I also miss. But as a hack you could use rooms as a secondary group mechanism since you don’t need them to represent actual rooms.

I am suspecting the problems you have are due to the name of the scene. It appears certain keywords do overrule locally-defined names. Whenever I put an address book entry for a taxi company with the name "taxi" in it, Siri isnt able to call it because it assumes I want to search for a taxi nearby. A similar thing might be going on with "apartment" and homekit. Frankly, a bit cynical, but by now, I am suspecting some intern put a rule in which contains "apartment" and has a very high priority, or somesuch...

Not directly related to Siri, but VoiceOver icon labels are also getting worse and worse over the years. For example, the weather app. In the beginning, when you scrolled with the VoiceOver cursor through the hourly forecast, it would say "Sonnenaufgang" and "Sonnenuntergang" (Sunrise and Sunset) at the appropriate places. Starting with iOS 12 (or was it 13?) the spoken label changed to "Sonne aufwärtspfeil" (sun arrow up). So someone went in there, decided that the nice description from earlier wasn't good enough, and replaced it with a literal description of the icon, ignoring context altogether.

> “I am suspecting the problems you have are due to the name of the scene.”

yah, for siri, i’ve found simple nouns (apartment) better than phrases (apartment lights off) for naming things in the home app, and for “all lights”, you don’t even need to create a scene (as others have pointed out). for rooms, you can just say “turn kitchen on” and it will turn all lights in its kitchen room on (which may or may not map to all lights in your actual kitchen, depending on how you set it up).

oh, and siri will refuse to do something if you curse at it: “hey siri, turn on the lights, you asshole” ominously responds with “sorry, i can’t do that (…dave)”

Except I don’t have this issue with Google Assistant. It guesses the correct word/phrasing nearly every time. Siri, OTOH, fails much more than it succeeds.

How has Doonesbury not put Siri out of her misery?

I started using this garbage with the 4S and while there have been under the hood improvements of onboard processing, etc, the core product has never gained the implicit context awareness promised and needed. As far as I can tell, it's a checkbox zombie "feature" that never met the usability threshold to become something for normal people but organizationally aged into slow development maturity.

https://www.computerhistory.org/revolution/mobile-computing/...

I’ll bite.

Me to watch: Hey Siri, call so-and-so on Bluetooth

Siri: calls so-and-so with Bluetooth selected but nonetheless the audio comes out the watch until I touch the phone.

Siri on Apple Watch is generally very weak.

Recently I went through the trouble to create lists to use Siri to add shopping items. I say, “add mustard oil”. Siri then adds “mustard” and “oil”. No problem, I’m smart. So I create a custom word in my phone and train it on my pronunciation. Try again, “mustard” and “oil”.

Guess I’m going to be shopping for “oil of mustard” then

I have also found Siri’s voice recognition for shopping list items lackluster, especially compared to Alexa. I’m guessing it has to do with Amazon having a better model of actual subjects/products due to having a storefront with said items.

It’s really bad when I’m at the store looking at something in my list scratching my head over what “Silicon Matt to sit rock song” is supposed to be without any way to hear what was actually said or give feedback that what Siri did was wrong.

"Hey Siri, is my phone still charging?" - "The current volume is 43%".

Siri can't start an outdoor walk, silly. She doesn't have legs.

Are you Scottish by any chance? https://www.youtube.com/watch?v=NMS2VnDveP8

I am not :/

Having multiple things actively listening for "Hey Siri" seems like a problem you've created, honestly.

I have my phone set to only engage Siri with the long button press. It saves a lot of trouble.

Blaming the user for the lack of simple coordination between devices when they buy into your ecosystem across multiple devices and attempt to use the advertised features doesn't seem like a great product strategy.

Is it really blame if it's simply a technical deficiency?

I imagine that if I put 10 iPhones on a desk and say "Hey Siri", that wouldn't work out too well. Doesn't mean we have to start "blaming" people, companies and things.

The phone and watch are advertised as intergraded with each other so this does not happen. Just like answering FaceTime on one device is supposed to stop the others from ringing. It is a feature of the watch and iPhone to figure out which should be listening and it is broken.

To the article’s point, Amazon’s has no problem Determining which should answer a question if they are all the same account piled on your desk.

And yet there are alarms set on the watch which are separate from alarms set on the paired phone. It’s really madness.

We are not talking about ten phones on a desk. We are talking about a phone and a watch, which are explicitly designed to and advertised as working together. Surely you see the distinction between your thought experiment and the actual scenario in question?

Surely you understand that the scenario is merely an attempt at illustrating the difference between blame and technical deficiency.

The user understands this limitation, and rather than lighting a candle is cursing the darkness.

It's redundant and counterproductive to have multiple devices listening for the code phrase.

the entire point of these assistant technologies is supposed to be that they seamlessly integrate with the environment. If you have to do long button presses or avoid this or that you can just open Spotify yourself and click play. Given that the benefit you get out of this stuff is likely marginal to begin with failing in 5% of cases is already way too much.

At this point to me it seems like most of it is basically toy technology without any real benefit unless you're visually impaired or unable to use your hands or something. People went from voice messaging to texting with other people because tactile inputs are often faster, so it's always puzzled me what the point is of talking to machines who can't even understand things properly half the time.

>the entire point of these assistant technologies is supposed to be that they seamlessly integrate with the environment

No. The "entire point" is to provide an easier way to do some interactions. On this point, they work super well. They're just not magic or perfect.

However, it definitely IS useful to state the point that way, because it makes it easier to argue that they're pointless toy technologies, thereby allowing you to feel superior.

Multiple devices are supposed to coordinate amongst themselves to decide who takes the request:

https://support.apple.com/en-us/HT208472

Reply on news.ycombinator.com